User Guide to

SIMCA


By MKS Umetrics
Version 13



1992-2012 MKS Umetrics AB, all rights reserved
Information in this document is subject to change without notice and does not represent
a commitment on part of MKS Umetrics. The software, which includes information
contained in any databases, described in this document is furnished under license
agreement or nondisclosure agreement and may be used or copied only in accordance
with the terms of the agreement. It is against the law to copy the software except as
specifically allowed in the license or nondisclosure agreement. No part of this user
guide may be reproduced or transmitted in any form or by any means, electronic or
mechanical, including photocopying and recording, for any purpose, without the
express written permission of MKS Umetrics.
Umetrics patents and trade marks:
OPLS/O2PLS, SWE-9802229-6, USA-6754543,
SEQUENTIAL MODELING, USA-7523384
OSC, SWE-0000563-7, USA-6853923
DESIGN SPACE USA-8086327
PLS-TREE, Pending
RED - MUP, VALUE FROM DATA, OPLS, O2PLS, O2-PLS, OPLS-DA,
O2PLS-DA
PLS-TREE, S-PLOT, EZinfo, SBOL, FABSTAT, BATCH
FINGERPRINT, SIMCA.

ID #2043
User guide edition date: April 2, 2012







MKS Umetrics AB
Stortorget 21
SE-211 34 Malm
Sweden
Phone: +46 (0)40 664 2580
Email: [email protected]


Welcome

Welcome to the user guide for SIMCA 13. This is your guide to SIMCA and its
capabilities.
This user guide does not include in-depth background material to multivariate data
analysis. For such information, see the book Multi- and Megavariate Data Analysis.
Content
The help is parted into 16 chapters, two appendices, and a reference list.
1. How to get started
2. Introduction to multivariate data analysis
3. Overview of SIMCA
4. Quick Access Toolbar
5. File
6. SIMCA import
7. Home
8. Data
9. Batch
10. Analyze
11. Predict
12. Plot/List
13. View
14. Plot and list contextual tabs
15. SIMCA-P+ 12 features - where to find them
16. Help
17. Appendix A: Statistics
18. Appendix B: Preprocessing
19. References


Assumed prior knowledge
We assume that you are already familiar with Windows. You should be familiar with
the topics discussed in your Microsoft Windows User's Guide, including how to use
the mouse to select, click, shift-click, control-click, press, drag, and choose from a
menu or tab. You should also know how to search through folders to find files.



v
Contents
How to get started 1
Introduction.................................................................................................................... 1
License information ....................................................................................................... 1
What is in the package ................................................................................................... 1
System requirements ...................................................................................................... 2
Installation ..................................................................................................................... 2
Registration and activation .................................................................................................. 2
Starting SIMCA ............................................................................................................. 2
SIMCA projects ............................................................................................................. 2
Work process for regular projects .................................................................................. 3
Creating a new project by importing a dataset .................................................................... 4
Viewing and preprocessing the data .................................................................................... 4
Specifying the workset ........................................................................................................ 5
Specifying and fitting the model ......................................................................................... 6
Reviewing the fit and performing diagnostics ..................................................................... 7
Using the model for predictions .......................................................................................... 7
Work process for batch modeling .................................................................................. 8
Batch project overview...................................................................................................... 10
Introduction to multivariate data analysis 13
Introduction.................................................................................................................. 13
Process data analysis .................................................................................................... 13
Process data properties ...................................................................................................... 14
Batch data analysis ...................................................................................................... 15
Batch data properties ......................................................................................................... 16
Reorganization of 3D table before import ......................................................................... 16
Traditional data analysis: One or two variables at a time ............................................ 17
Multivariate approach .................................................................................................. 18
Summarizing a set of process variables ............................................................................. 18
Summarizing batch data .................................................................................................... 18
PCA - Principal Component Analysis ............................................................................... 19
PLS - Partial Least Squares Projections to Latent Structures ............................................ 19
OPLS and O2PLS - Orthogonal Partial Least Squares ...................................................... 20
Batch modeling in SIMCA ................................................................................................ 20
Conclusion ................................................................................................................... 26
Overview of SIMCA 27
Overview of SIMCA .................................................................................................... 27
Application icon and symbol ....................................................................................... 27
Projects ........................................................................................................................ 27
Datasets ............................................................................................................................. 28
Workset ............................................................................................................................. 28
Models .............................................................................................................................. 28
SIMCA window ........................................................................................................... 29
SIMCA ribbon description ........................................................................................... 30
User Guide to SIMCA
vi
License information .......................................................................................................... 30
Ribbon ......................................................................................................................... 31
Quick Access Toolbar ....................................................................................................... 31
File tab .............................................................................................................................. 31
Home ................................................................................................................................. 32
Data ................................................................................................................................... 33
Batch ................................................................................................................................. 34
Analyze ............................................................................................................................. 35
Predict ............................................................................................................................... 36
Plot/List ............................................................................................................................. 36
View .................................................................................................................................. 37
Tools contextual tab .......................................................................................................... 38
Layout contextual tab ........................................................................................................ 39
Marked Items contextual tab ............................................................................................. 40
Help ................................................................................................................................... 41
Shortcut menu .............................................................................................................. 41
Conventions ................................................................................................................. 41
Limitations in project names ............................................................................................. 41
Limitations in observation and variable identifiers ........................................................... 41
Case sensitivity ................................................................................................................. 41
Menu and tab reference syntax .......................................................................................... 41
Select and mark ................................................................................................................. 42
Vector and matrix representation ...................................................................................... 42
Missing values representation ........................................................................................... 43
Presentation of SIMCA results .................................................................................... 44
Plots for publications ......................................................................................................... 44
Plots for reports ................................................................................................................. 44
Quick Access Toolbar 45
Quick Access Toolbar .................................................................................................. 45
Customizing Quick Access Toolbar ............................................................................ 45
Removing buttons from Quick Access Toolbar ................................................................ 45
Adding buttons to Quick Access Toolbar .......................................................................... 46
Moving the Quick Access Toolbar .................................................................................... 46
Minimizing the ribbon ................................................................................................. 46
Find .............................................................................................................................. 47
Marking type ..................................................................................................................... 48
Using Find ......................................................................................................................... 49
File 51
Introduction.................................................................................................................. 51
New - Creating a new project ...................................................................................... 51
Open - Opening an existing project ............................................................................. 52
Opening pre-13 projects .................................................................................................... 52
Opening pre-13 filtered projects limitations ...................................................................... 52
Opening pre-13 batch projects .......................................................................................... 52
Save project ................................................................................................................. 52
Save As ........................................................................................................................ 52
Save Project As ................................................................................................................. 53
Save Plot or List As ........................................................................................................... 53
Print, Print Preview, and Print Setup ........................................................................... 54
Manage ........................................................................................................................ 54
Table Of Contents
vii
Encrypt .............................................................................................................................. 54
Compact Project File ......................................................................................................... 54
Delete Project .................................................................................................................... 55
Import .......................................................................................................................... 55
Generate Report ........................................................................................................... 55
Introduction ....................................................................................................................... 55
Starting to generate a report .............................................................................................. 56
Generate Report window ................................................................................................... 56
Default templates .............................................................................................................. 57
Appending to, inserting in, or replacing existing report .................................................... 58
Creating a report including several phases ........................................................................ 58
Menu bar in report ............................................................................................................. 59
Generate report bar ............................................................................................................ 63
Format bar ......................................................................................................................... 64
Placeholders window ........................................................................................................ 64
Properties window ............................................................................................................. 67
Adding plots and lists to the report.................................................................................... 68
Close project ................................................................................................................ 68
Recent files list ............................................................................................................ 68
Project Options ............................................................................................................ 69
Fitting options ................................................................................................................... 69
Predictions presentation .................................................................................................... 71
Distance to the model plot options .................................................................................... 71
Selecting plot labels .......................................................................................................... 72
Selecting type of residuals and R2 .................................................................................... 73
Performance improvement using data caching .................................................................. 74
Administration of the audit trail ........................................................................................ 74
General .............................................................................................................................. 76
SIMCA Options ........................................................................................................... 77
General .............................................................................................................................. 77
Selecting multithreading ................................................................................................... 78
Restoring to default ........................................................................................................... 79
Selecting progress bar pictures .......................................................................................... 80
More Options .................................................................................................................... 80
Exit............................................................................................................................... 82
SIMCA import 83
Introduction.................................................................................................................. 83
File ............................................................................................................................... 84
New Spreadsheet ............................................................................................................... 84
Workspace ......................................................................................................................... 96
Save As in SIMCA import ................................................................................................ 96
Print, Print Preview, and Print Setup ................................................................................. 96
Close ................................................................................................................................. 96
Options .............................................................................................................................. 97
Quick Access Toolbar .................................................................................................. 98
Find toolbar .................................................................................................................. 98
Home tab .................................................................................................................... 100
Specifying identifiers ...................................................................................................... 100
Excluding/including rows or columns ............................................................................. 105
Specifying data properties ............................................................................................... 105
Column and row menus................................................................................................... 108
User Guide to SIMCA
viii
Importing batch conditions .............................................................................................. 109
Finish and create project ................................................................................................. 109
Edit tab ....................................................................................................................... 110
Creating observation or variable index ............................................................................ 110
Rows and Columns ......................................................................................................... 110
Merge .............................................................................................................................. 111
Apply Formatting ............................................................................................................ 113
Sort .................................................................................................................................. 113
View tab ..................................................................................................................... 113
Issues pane in SIMCA import ......................................................................................... 113
Audit trail ........................................................................................................................ 118
Variables pane ................................................................................................................. 118
Observations pane ........................................................................................................... 119
Class pane ....................................................................................................................... 119
Batch & Phase pane phase part ....................................................................................... 119
Batch & Phase pane batch part ........................................................................................ 121
Summary pane ................................................................................................................. 122
Displaying the Missing Value Map ................................................................................. 122
Home 123
Introduction................................................................................................................ 123
Dataset spreadsheet .................................................................................................... 123
Shortcut menu in the dataset spreadsheet ........................................................................ 123
Adding observations and variables .................................................................................. 124
Adding secondary IDs ..................................................................................................... 125
Workset ...................................................................................................................... 125
Content ............................................................................................................................ 125
Workset Statistics ............................................................................................................ 126
Correlation Matrix ........................................................................................................... 127
New workset ................................................................................................................... 128
New workset as model .................................................................................................... 129
Editing the workset ......................................................................................................... 130
Deleting the workset ....................................................................................................... 131
Workset dialog ................................................................................................................ 131
Simple mode workset wizard .......................................................................................... 168
Change Model Type ........................................................................................................ 171
Model Options ................................................................................................................. 177
Fit model .................................................................................................................... 186
Autofit ............................................................................................................................. 186
Two First Components .................................................................................................... 187
Add component ............................................................................................................... 187
Remove component ......................................................................................................... 188
Diagnostics & Interpretation ...................................................................................... 188
Overview button in the Diagnostics & Interpretation group ........................................... 188
Summary of Fit ............................................................................................................... 189
Scores .............................................................................................................................. 195
Loadings .......................................................................................................................... 204
Hotellings T2Range ....................................................................................................... 213
Distance to Model ........................................................................................................... 214
Observed vs. Predicted .................................................................................................... 216
Coefficients ..................................................................................................................... 216
VIP .................................................................................................................................. 220
Table Of Contents
ix
Data 221
Introduction................................................................................................................ 221
Merge ......................................................................................................................... 221
Split dataset ................................................................................................................ 222
Transpose ................................................................................................................... 223
Delete Dataset ............................................................................................................ 224
Generate Variables ..................................................................................................... 224
Generate Variables wizard .............................................................................................. 224
User generated functions ................................................................................................. 226
New variables from other dataset .................................................................................... 227
Generating variables in batch projects ............................................................................ 227
Local Centering ......................................................................................................... 228
Importing local centering ................................................................................................ 229
Local centering of the predictionset ................................................................................ 230
Local centering missing .................................................................................................. 230
View imported local centering ........................................................................................ 230
Spectral Filters ........................................................................................................... 231
Filtering limitations ......................................................................................................... 231
First, Second, and Third Derivatives ............................................................................... 232
Multiplicative Signal Correction MSC ......................................................................... 234
Standard Normal Variate SNV ..................................................................................... 235
Row Center ..................................................................................................................... 236
Savitzky-Golay ................................................................................................................ 237
EWMA ............................................................................................................................ 239
Wavelet Compression Spectral WCS ........................................................................... 241
Wavelet Denoise Spectral WDS ................................................................................... 247
Orthogonal Signal Correction OSC .............................................................................. 251
Chaining filters ................................................................................................................ 255
Predictionsets and filtered datasets .................................................................................. 255
Time Series Filters ..................................................................................................... 256
Wavelet Compress Time Series WCTS ........................................................................ 256
Wavelet Denoising/Decimation WDTS ....................................................................... 261
Dataset Summary ....................................................................................................... 266
General page in dataset Properties .................................................................................. 267
Observations page ........................................................................................................... 267
Variables page ................................................................................................................. 268
Phases .............................................................................................................................. 269
Batches ............................................................................................................................ 270
Filter summary ................................................................................................................ 270
Trimming Overview .................................................................................................. 271
Displaying the Missing Value Map ........................................................................... 272
Spectra ....................................................................................................................... 272
Hierarchical models ................................................................................................... 273
Hierarchical models with OPLS or O2PLS ..................................................................... 274
Hierarchical models with blocks ..................................................................................... 274
Hierarchical base models reverts to non hierarchical ...................................................... 274
Batch 275
Introduction................................................................................................................ 275
Batch Control Charts ................................................................................................. 275
Batch control chart background ...................................................................................... 276
User Guide to SIMCA
x
Batch control chart plot and list customization ............................................................... 276
Excluding batches ........................................................................................................... 279
Score Batch Control Chart .............................................................................................. 280
Variable Batch Control Chart .......................................................................................... 280
DModX Batch Control Chart .......................................................................................... 281
Hotellings T2Range Batch Control Chart ...................................................................... 281
Observed vs. Time/Maturity Batch Control Chart .......................................................... 282
Batch Control Charts List ................................................................................................ 282
Batch control charts for new batches ......................................................................... 284
Create BLM ............................................................................................................... 285
Selecting what to base the batch level dataset on ............................................................ 286
Selecting BEM variables for BL DS dialog .................................................................... 287
Raw Data Statistics Types dialog .................................................................................... 288
Batch condition datasets .................................................................................................. 289
Batch level datasets and missing values .......................................................................... 289
Sources of Variation plot ........................................................................................... 290
BLM with batch conditions ............................................................................................. 290
BLM loading and contribution plots example ................................................................. 291
Time/Maturity group plots ......................................................................................... 292
Observed vs. Smoothed Y plot ........................................................................................ 292
Unaligned vs. Aligned plot .............................................................................................. 293
Creating hierarchical models for batch level datasets ................................................ 294
Create Hierarchical Batch Models dialog ........................................................................ 294
Create Hierarchical Batch Models results ....................................................................... 296
Hierarchical top level model ........................................................................................... 298
Batch Variable Importance Plot ................................................................................. 299
Analyze 301
Introduction................................................................................................................ 301
Biplot ......................................................................................................................... 302
Colors in the Biplot ......................................................................................................... 302
Inner Relation Plot ..................................................................................................... 303
S-plots ........................................................................................................................ 303
S-plot ............................................................................................................................... 303
S-line plot ........................................................................................................................ 304
SUS-Plot ......................................................................................................................... 305
Contribution plots ...................................................................................................... 305
Scores/T2 Contribution plot ............................................................................................ 306
Distance to Model X Contribution plot ........................................................................... 308
Y Predicted Contribution plot ......................................................................................... 311
Contribution plot for hierarchical top level models ......................................................... 313
Combined Contribution plot for batch level model ......................................................... 313
Contribution plot from plot ............................................................................................. 314
Coefficient Y-Related Profiles plot ........................................................................... 316
RMSEcv plot ............................................................................................................. 317
Residual normal probability plot ............................................................................... 318
Permutation plot ......................................................................................................... 319
CV-ANOVA .............................................................................................................. 320
CV Scores plot ........................................................................................................... 320
HCA ........................................................................................................................... 321
Clustering of observations (matrix rows) ........................................................................ 321
Table Of Contents
xi
Clustering of variables (matrix columns) ........................................................................ 324
HCA options ................................................................................................................... 325
Group Dendrogram Clusters marking tool ...................................................................... 325
Creating a PLS-DA or OPLS-DA model from plot marking .......................................... 326
Creating class models from plot marking ........................................................................ 326
PLS-Tree .................................................................................................................... 326
Initializing a PLS-Tree .................................................................................................... 326
PLS-Tree wizard - removing outliers .............................................................................. 327
Clustering algorithm parameters ..................................................................................... 327
PLS-Tree resulting dendrogram ...................................................................................... 328
Group Dendrogram Clusters marking tool ...................................................................... 329
PLS-Tree model window and partial models .................................................................. 330
Predict 331
Introduction................................................................................................................ 331
Specify Predictionset ................................................................................................. 332
As Dataset ....................................................................................................................... 334
As Workset ...................................................................................................................... 334
Complement Workset ...................................................................................................... 334
Class ................................................................................................................................ 334
Delete Predictionset ........................................................................................................ 334
Batch predictionset .......................................................................................................... 335
Predictionset for filtered datasets .................................................................................... 335
Missing values in the predictionset ................................................................................. 335
Prediction List ............................................................................................................ 335
Properties in Prediction List ............................................................................................ 336
Y PS plots .................................................................................................................. 337
RMSEP ........................................................................................................................... 337
Y PS Scatter Plot ............................................................................................................. 338
Y PS Line Plot ................................................................................................................. 338
Y PS Column Plot ........................................................................................................... 339
Y Predicted List .............................................................................................................. 339
Score PS plots ............................................................................................................ 340
Predicted Score Scatter Plot ............................................................................................ 340
Predicted Score Line Plot ................................................................................................ 340
Predicted Score Column Plot .......................................................................................... 341
Predicted Score 3D Plot .................................................................................................. 341
Hotellings T2Range plot ........................................................................................... 342
DModX PS ................................................................................................................ 342
DModX PS - Distance to Model X-Block ....................................................................... 343
DModY PS - Distance to Model Y-Block ....................................................................... 343
Control Charts PS ...................................................................................................... 344
Contribution PS ......................................................................................................... 345
Scores/T2 ........................................................................................................................ 345
Distance to Model X-Block Contribution plot ................................................................ 346
Distance to Model Y-Block Contribution plot ................................................................ 346
Y Predicted Contribution plot ......................................................................................... 347
Contribution plot from plot ............................................................................................. 348
Time Series PS ........................................................................................................... 349
Time Series plot example ................................................................................................ 349
Coomans' Plot ............................................................................................................ 350
Classification List for class models ........................................................................... 351
User Guide to SIMCA
xii
Coloring in the Classification List ................................................................................... 351
Classification List for Discriminant Analysis models ............................................... 352
Coloring in the Classification List ................................................................................... 352
Misclassification Table .............................................................................................. 352
Misclassification Table for class models ......................................................................... 353
Misclassification Table - Discriminant Analysis models ................................................ 354
Misclassification Table Properties .................................................................................. 354
Plot/List 357
Introduction................................................................................................................ 357
Name conventions ..................................................................................................... 357
Standard plot types and list dialog ............................................................................. 358
Data Series page .............................................................................................................. 359
Transformation ................................................................................................................ 361
Control Charts ............................................................................................................ 368
What is a control chart? ................................................................................................... 368
Control Chart dialog ........................................................................................................ 369
Response Contour ...................................................................................................... 379
Response Contour plot options ........................................................................................ 380
Response Surface Plot ............................................................................................... 381
Wavelet Structure ...................................................................................................... 381
Displaying Wavelet Structure plot .................................................................................. 382
Wavelet Power Spectrum ........................................................................................... 382
Displaying Wavelet Power Spectrum plot ....................................................................... 383
Wavelet Power Spectrum example .................................................................................. 383
Step Response Plot .................................................................................................... 384
Creating a finite impulse response model ........................................................................ 384
Displaying Step Response Plot ........................................................................................ 386
Using an existing finite impulse response model ............................................................ 387
View 389
Introduction................................................................................................................ 389
Show/Hide ................................................................................................................. 389
Audit Trail ....................................................................................................................... 390
Analysis Advisor ............................................................................................................. 392
Favorites .......................................................................................................................... 392
Item Information ............................................................................................................. 395
Marked Items .................................................................................................................. 396
Model window ................................................................................................................ 398
Observations window ...................................................................................................... 402
Project window ............................................................................................................... 402
Quick Info ....................................................................................................................... 405
Variables window............................................................................................................ 421
Full Screen ...................................................................................................................... 421
Add to Favorites ........................................................................................................ 421
Add to Report ............................................................................................................ 421
Window group ........................................................................................................... 421
Plot and list contextual tabs 423
Mini-toolbar - format plot .......................................................................................... 424
Managing the mini-toolbar .............................................................................................. 425
Table Of Contents
xiii
Tools tab .................................................................................................................... 426
Change Type ................................................................................................................... 426
Data group ....................................................................................................................... 432
Properties group .............................................................................................................. 433
Select - Marking tool ....................................................................................................... 444
Selecting points according to color in legend .................................................................. 446
Find ................................................................................................................................. 446
Highlight Series ............................................................................................................... 448
Sorting ascending or descending ..................................................................................... 448
Zooming in ...................................................................................................................... 450
Zooming out .................................................................................................................... 450
Layout tab .................................................................................................................. 450
Templates group .............................................................................................................. 451
Show group ..................................................................................................................... 471
Marked Items tab ....................................................................................................... 472
Create from Marked Items group .................................................................................... 472
Drill Down group ............................................................................................................ 474
Modify model .................................................................................................................. 479
Layout group ................................................................................................................... 482
SIMCA-P+ 12 features - where to find them 485
Introduction................................................................................................................ 485
SIMCA-P+ 12 File menu ........................................................................................... 486
Import wizard .................................................................................................................. 489
SIMCA-P+ 12 Edit menu .......................................................................................... 498
SIMCA-P+ 12 View menu ........................................................................................ 498
SIMCA-P+ 12 View menu - Toolbars ............................................................................. 500
SIMCA-P+ 12 Dataset menu ..................................................................................... 503
SIMCA-P+ 12 Workset menu .................................................................................... 504
SIMCA-P+ 12 Analysis menu ................................................................................... 504
SIMCA-P+ 12 Predictions menu ............................................................................... 506
SIMCA-P+ 12 Plot/List menu ................................................................................... 508
SIMCA-P+ 12 Window menu ................................................................................... 509
SIMCA-P+ 12 Help menu ......................................................................................... 510
SIMCA-P+ 12 shortcut menu .................................................................................... 510
SIMCA-P+ 12 shortcut menu - Create ............................................................................ 512
SIMCA-P+ 12 Color in Properties dialog ....................................................................... 513
SIMCA-P+ 12 Properties dialog ..................................................................................... 514
SIMCA-P+ 12 Plot Settings ............................................................................................ 514
Help 521
Introduction................................................................................................................ 521
HTML help ................................................................................................................ 521
Register ...................................................................................................................... 521
Umetrics on the Web ................................................................................................. 521
About SIMCA ............................................................................................................ 522
Appendix A: Statistics 523
Introduction................................................................................................................ 523
Fit methods background ............................................................................................. 523
PCA - Principal Components modeling .......................................................................... 524
User Guide to SIMCA
xiv
PLS - Partial Least Squares Projection to Latent Structures modeling ............................ 525
OPLS/O2PLS - Orthogonal PLS modeling ..................................................................... 526
Cluster Analysis (CA), dendrograms, Hierarchical CA (HCA), PLS-Tree ............... 530
Hierarchical Cluster Analysis - HCA .............................................................................. 530
PLS-Trees ....................................................................................................................... 531
References cluster analysis .............................................................................................. 533
Vectors available in SIMCA ...................................................................................... 533
Variables and Scores vectors ........................................................................................... 534
Observations and Loadings vectors ................................................................................. 541
Function of component vectors ....................................................................................... 548
Function of lags vectors .................................................................................................. 551
Aligned Vectors .............................................................................................................. 551
Batch Vectors .................................................................................................................. 553
Formulas and descriptions ......................................................................................... 555
Prediction vectors ............................................................................................................ 555
R2Y, R2X, R2Yadj, R2Xadj ........................................................................................... 555
R2V and R2Vadj ............................................................................................................. 556
Q2 and Q2V .................................................................................................................... 556
Q2(cum) and Q2V(cum) ................................................................................................. 556
Leverages ........................................................................................................................ 557
RSD of observations and variables.................................................................................. 557
Relevance of variables .................................................................................................... 557
Hotelling's T2 .................................................................................................................. 557
Missing values correction factor ..................................................................................... 558
Missing values and PCA ................................................................................................. 558
Score and loading vectors ............................................................................................... 559
T scores (X): t1 vs. t2, ... ................................................................................................. 559
U scores (Y): u1 vs. u2, ... ............................................................................................... 559
T vs. U scores (X&Y): t1 vs. u1, ... ................................................................................. 559
P loadings (X): p1 vs. p2,... ............................................................................................. 559
W loadings (X): w1 vs. w2, ... ......................................................................................... 559
W* loadings (X): w*1 vs. w*2, ... ................................................................................... 560
C loadings (Y): c1 vs. c2, ... ............................................................................................ 560
WC or W*C loadings (X&Y): wc1 vs. wc2, ... ............................................................... 560
TPS scores (X): tPS1 vs. tPS2, ... .................................................................................... 560
Distance to the model ...................................................................................................... 560
Variable importance, VIP ................................................................................................ 564
Standard Error and Jack-knifing uncertainties ................................................................ 565
Coefficients ..................................................................................................................... 565
RMSEE, RMSEP and RMSECV .................................................................................... 567
Transform page criteria .............................................................................................. 568
Automatic transformation criteria ................................................................................... 568
Scaling ....................................................................................................................... 569
Autoscaling ..................................................................................................................... 569
Block scaling ................................................................................................................... 569
Scaling weight calculation .............................................................................................. 569
Scaling of expanded terms .............................................................................................. 570
Scaling weights for transformed variables ...................................................................... 570
Scaling of variables with classes ..................................................................................... 570
Scaling of lagged variables ............................................................................................. 570
Scaling after changing the observation selection ............................................................. 570
Cross validation ......................................................................................................... 571
Cross validation for PCA ................................................................................................ 571
Table Of Contents
xv
Cross validation for PLS, OPLS and O2PLS .................................................................. 571
Reference cross validation .............................................................................................. 572
Cross validation rules - Significant component ............................................................... 572
Cross validation rules - Non Significant component ....................................................... 573
Cross validation rules for batches ................................................................................... 573
PLS time series analysis ............................................................................................ 574
References PLS time series ............................................................................................. 575
CV-ANOVA .............................................................................................................. 575
The ANOVA table .......................................................................................................... 576
References CV-ANOVA ................................................................................................. 577
Fisher's Exact test ...................................................................................................... 577
References Fisher's Exact test ......................................................................................... 577
Control Chart statistics ............................................................................................... 578
Nomenclature and notation in Control Charts ................................................................. 578
Target and Standard deviation ......................................................................................... 579
S-plot ......................................................................................................................... 579
Reference S-plot .............................................................................................................. 579
Observation Risk ....................................................................................................... 580
Appendix B: Preprocessing 581
Preprocessing available in SIMCA ............................................................................ 581
Auto and cross correlation of variables or observations ............................................ 582
Reconstruction of wavelet compressed data .................................................................... 582
References auto and cross correlation ............................................................................. 583
Power Spectrum Density ........................................................................................... 583
Reference power spectrum .............................................................................................. 583
Wavelet transformations ............................................................................................ 583
Introduction ..................................................................................................................... 583
Overview of the wavelet transform ................................................................................. 584
Wavelet theory ................................................................................................................ 584
Multiresolution analysis, MRA ....................................................................................... 585
Wavelet families ............................................................................................................. 589
Wavelet compression or de-noising of signals ................................................................ 590
References for wavelets .................................................................................................. 593
Filtering and calibration ............................................................................................. 594
Derivatives ................................................................................................................. 594
Multiplicative Signal Correction (MSC) .................................................................... 595
References for MSC ........................................................................................................ 595
Standard Normal Variate (SNV) ................................................................................ 595
Reference for SNV .......................................................................................................... 596
Orthogonal Signal Correction (OSC) ......................................................................... 596
Reference for OSC .......................................................................................................... 597
References 599
References for multivariate analysis .......................................................................... 599
Glossary 621
Index 627

1
How to get started
Introduction
This chapter holds information about the installation, including system requirements
and such, as well as a short introduction to using SIMCA.
Content
- License information
- What is in the package
- System requirements
- Installation
- Starting SIMCA
- SIMCA projects
- Work process for regular projects
- Work process for batch modeling
License information
Before installing the SIMCA product, carefully read the license agreement included
with your SIMCA software package.
In case you do not fully accept the terms of the license agreement, you should
immediately return all parts of the package to your local supplier.
What is in the package
The SIMCA package contains the following:
- Installation files.
- SIMCA User Guide (.pdf).
- Installation instructions document.
User Guide to SIMCA
2
System requirements
The minimum recommended system requirements:
- Pentium based computer (PC) with a 1.5 GHz or faster processor.
- 1 GB RAM or more.
- 1 GB available hard disk space.
- Color graphics display with at least 800x600 resolution.
- DVD-ROM or CD-ROM drives if the SIMCA package was delivered on CD.
- Microsoft Windows XP, Vista, or 7.
- Graphics card that has hardware 3D acceleration and supports Open GL.
Installation
To install, follow the installation instructions delivered with the installation program.
Note: You must have administrative privileges to be able to install the
software.
Registration and activation
To register and activate, follow the instructions delivered with the installation program.
See also the Register section in the Help chapter.
Starting SIMCA
Start SIMCA by double clicking the icon or by starting it from the Start-menu.
Continue by:
- Reading about the SIMCA software in the Help or User Guide. Both contain
same information.
- Starting a new investigation: click File | New.
- Running tutorial examples: Find the tutorials and tutorial datasets at the
Umetrics web site www.umetrics.com, or contact your Umetrics sales
office. Select a suitable tutorial example, import the listed dataset used in the
tutorial and follow the described steps.
SIMCA projects
What is a project?
SIMCA is organized into projects. A project is a .usp file containing the results of the
analysis of one or more datasets.
Creating a new project
Start a new project by importing a dataset. The default unfitted model is created when
exiting the SIMCA import and it is displayed in the Project Window.
How to get started
3
Project window
The project window displays, for every model, one line summarizing the model results.
The active model, the one you are working with, is marked in the project window.
Default workset and model
The default Workset consists of all variables and observations imported with the roles
defined in the import. All variables are default centered and scaled to unit variance.
How the variables were defined at import also defines which type the default model is.
Creating a new model
Unfitted models are implicitly created by SIMCA when:
- Specifying a workset or
- When changing the model type of a fitted model (fitted model is the active
model).
Activating and opening the model window
To activate a model click it in the project window.
All plots and lists created from the Home, Analyze and Predict tabs are created for the
active model.
To open the model window:
- Double-click it in the project window or
- Mark the model in the project window and then select the Model Window
check box on the View tab.
The model window opens with the details of the model results, displaying one line per
component.
Opening the workset dialog
To open the workset dialog use one of the following methods:
1. Right-click the model in the Project Window and then select Edit Model x.
2. Click the Workset button in the Model Window.
3. On the Home tab click Edit | Model x.
Work process for regular projects
The work process in SIMCA consists of the following steps:
Step Objective How to do it
1. Creating a new project. Click File | New Regular Project, and then
select the dataset file or files to import. In the
import spreadsheet define identifiers, X, Y,
and qualitative variables etc.
2. Viewing and
preprocessing the
data.
When warranted, preprocess your dataset
using the available Spectral Filters or Time
Series Filters on the Data tab.
User Guide to SIMCA
4
Step Objective How to do it
3. Specifying the workset
and model type.
On the Home tab, open the Workset dialog
by clicking New/Edit in the Workset group.
Select the variables and observations to
include or exclude, define classes, transform,
scale, or lag variables, and Trim-Winsorize
the variables. Optionally select Model type
before clicking OK to exit the dialog.
4. Fitting the model. In the Fit model group click Autofit.
5. Reviewing the results
and
performing diagnostics.
Create plots found in the Diagnostics &
Interpretation group to detect trends,
groupings, deviations, etc.
6. Using the model for
predictions.
On the Predict tab, use the predictionset to
see how new data fit with the active model.
Import another dataset with the new
observations or use observations not
included in the model as predictionset.
Creating a new project by importing a dataset
Create a new project by importing a dataset and specifying data properties in the
SIMCA import spreadsheet.
Selecting the dataset
Click File | New Regular Project and then browse to find the file or files to import.
To import from a database, Cancel the Open dialog and then click New Spreadsheet |
From Database in the SIMCA import.
You can import more than one dataset at the same time in SIMCA import by clicking
File | New Spreadsheet. Selecting another file adds a new spreadsheet to the SIMCA
import. These spreadsheets are imported individually at Finish and you can create the
workset from a selection of the datasets.
Indicating file contents
Specify Primary ID and as many Secondary IDs as desired for both variables and
observations.
Specify qualitative and date/time variables when present.
Clicking Finish imports all currently open spreadsheets.
Viewing and preprocessing the data
Click the Dataset-button (on the Home tab) to open any available dataset. Use the
Data tab to manipulate and preprocess datasets.
Viewing the dataset
Before preprocessing your data you may want to display the variables or observations
in a plot. You can do this directly from the dataset spreadsheet by marking the
variables or observations and on the Marked Items tab clicking one of the plots in the
Create from Marked Items group, or XObs/YObs or by opening the Quick Info (on
the View tab).
How to get started
5
To plot all the X observations as line plot, click Data | Spectra | dataset.
Preprocessing the dataset
Quick Info
The Quick Info window holds interactive plots tied to the dataset displaying variables
or observations in the time or frequency domain. When displaying variables,
Trimming / Winsorizing for a single or all variables is available.
Generate new variables
Generate new variables as functions of existing ones or from model results (Data |
Generate Variables).
Filter the dataset
Filter your dataset (Data | Spectral Filters), for example using:
- Multiplicative Signal Correction (MSC).
- Standard Normal Variate (SNV).
- Wavelet transform and compression.
- Wavelet transform time series data using compression, decimation/denoising.
Specifying the workset
The default Workset, at the project start, is the first dataset with variables defined as X
and Y as specified at import, centered and scaled to unit variance. The associated
model (unfitted) is listed in the project window.
To fit the default model click Autofit in the Fit model group (Home tab).
An unfitted model is generated by SIMCA when clicking OK in the Workset dialog
and when switching model type for a fitted model by clicking the Change Model
Type-button.
New
In the Workset group click New to create a workset of the entire firstly imported dataset
with X and Y as defined at import and scaling as defined in Project Options on the File
tab.
New as model
In the Workset group click New as | Mx to use the workset of the selected model as
starting point.
Edit
In the Workset group click Edit to open an already created workset.
Modifying the workset
To fit a model different from the default model, with maybe other or more datasets
included, excluded variables and observations, or transformations, or different scaling,
it is necessary to modify the default Workset.
User Guide to SIMCA
6
The following pages are available as tabs in the Workset dialog:
Tab Functionality available
Select Data Selecting which datasets to include. Only available when there are
more than one dataset to choose from.
Variables Excluding and including variables. Defining variables as X or Y.
Observations Including and excluding observations. Grouping observations into
classes for classification.
Transform Applying transformations for variables. Trim-Winsorizing variables.
Note that trimming / Winsorizing the workset does not affect the
dataset but just that particular workset.
Lag Defining lag structure.
Expand Expanding terms with cross, square, and cubic terms.
Scale Scaling variables.
Spreadsheet Trim-Winsorizing variables. Note that trimming / Winsorizing the
workset does not affect the dataset but just that particular workset.
Simple mode
You can select to have the workset wizard guide you through the specification and
automatically fitting of the model by clicking the Use Simple Mode button in the
Workset dialog. To turn it off, click the Use Advanced Mode button.
Specifying and fitting the model
Selecting the model type
You can change the model type, and when the workset specification allows it, by selecting
another fit method than the default in the Model type box in the Workset dialog or by
clicking Change Model Type in the Workset group.
Selecting a new model type after fit
You can, after fitting the model, select a new model type. SIMCA then creates a new
unfitted model with the selected model type.
For example, if you have defined your Workset variables as X and Y, you can first fit a
PCY (PC of the responses), then change the model type to PLS and fit a PLS
model (another model) to the same data.
Fitting the model
Fit the model using one of the fitting types listed.
Autofit
Rule based fitting of the marked model or group of models.
Two First
Calculates two components whether they are significant or not. Often used to get a
quick overview of the data.
How to get started
7
Add
Calculate one component at a time. Here it is possible to force components to be
calculated regardless of significance rules.
Remove
Remove the last component
Autofitting class models
To fit a group of models in a wrapper model, CM, mark the wrapper and click Autofit.
All models in the wrapper are then autofitted.
Reviewing the fit and performing diagnostics
After fitting, the whole spectrum of plots and lists are available for model
interpretation and diagnostics.
Important plots
- Summary plots
- Scores: t1 vs. t2, t1 vs. u1, etc.
- Loadings: p1 vs. p2, w*c1 vs. w*c2, etc.
- Hotelling's T2Range
- Coefficients
- Variable Importance (VIP)
Important plots for diagnostics
- Distance to Model
- Residuals N-Plot (Analyze tab)
- Contribution plots
- Permutations (Analyze tab)
- For batch evolution models: batch control charts (Batch tab)
Note: By default, all cumulative plots and lists are displayed for the last
component. To select a different component for display in the plots, and/or a
different variable, introduce the change in the Data group, on the Tools tab.
Using the model for predictions
Importing datasets for prediction
Later you may import additional data for use in predictions. Click File | Import
Dataset.
Building the predictionset
Click the Specify-button on the Predict tab to build your predictionset from any or all
datasets. You can display the predictionset as a spreadsheet or just plot or list results.
If you do not specify a predictionset, the predictionset is by default the first dataset.
User Guide to SIMCA
8
Displaying the predictions
Use any fitted model to:
- Classify the observations from the predictionset with respect to a PC or PLS
model.
- Predict responses (values of Y variables) for the observations in the
predictionset with respect to a PLS or OPLS/O2PLS model.
All the prediction results (scores, y-values, etc.), computed with the active model, can
be displayed as plots or lists.
Plots/Lists
On the Plot/List tab you can find general plot and list routines. Here it is possible to
plot and list any data, results from the analysis and predictions.
There are scatter, line, column, 3d scatter, list, time series, control charts,
histogram, normal probability plots, response contour, response surface, wavelets
plots, and step response plots available.
Note: With an active plot or list, use the Change Type group, on the Tools
tab, to generate lists from plots and plots from lists.
Work process for batch modeling
A SIMCA batch project consists of one or more batch evolution datasets and models
and optionally one or more batch level datasets and models:
1. The batch evolution datasets with several observations per batch with the
variables measured during the evolution of the batch.
2. The batch level datasets consisting of the completed batches, with one batch
being one observation (matrix row). The variables of the batch level datasets
are the scores, original variables (raw), or statistical summary variables (raw
data statistics) of the batch evolution model, BEM, at every time point folded
out side-wise. The BLM may also contain batch condition variables.
Batches may be divided into phases.
How to get started
9
The work process in SIMCA consists of the following steps:
Step Objective How to do it
1. Creating a new project. Click File | New Batch Project
(alternatively File | New Regular Project |
New Batch Project if the last created
project was regular). Select the dataset file
or files to import. In the SIMCA import
spreadsheet, define Batch ID, Phase ID,
X, qualitative, time or maturity as Y, and if
present Batch Conditions.
With phases, in the Batch window
optionally Rename, Merge, Delete, switch
the order or time/maturity for the phases,
or apply Conditional Delete.
Batch conditions can be imported with the
batch evolution dataset, or as a separate
dataset. SIMCA import assumes that a
dataset with only one row per batch
contains batch conditions.
2. Viewing and
preprocessing the data.
When warranted, preprocess your dataset
using the available Spectral Filters or
Time Series Filters on the Data tab.
3. Specifying the workset. On the Home tab, open the Workset
dialog by clicking New/Edit in the Workset
group. Select the variables and
observations to include or exclude, crop,
transform, scale, or lag variables, and
Trim-Winsorize the variables. Use the
Variables page to configure the x and y-
variables to the relevant phases, and y-
variables smoothing/shifting/normalizing.
4. Fitting the model. Mark the wrapper BEM and click Autofit.
5. Reviewing the batch
evolution results and
performing diagnostics.
Create BCC-plots found in the Analysis
Control Charts group on the Batch tab to
detect trends, groupings, deviations, etc.
6. Using the model for
predictions.
Create the BCC-plots found in the
Prediction Control Charts group to see
how new data fit with the active model.
7. Creating the batch level
dataset.
On the Batch tab, click Create Batch
Level and go through the wizard.
8. Importing batch
conditions.
To import batch conditions, click File |
Import Dataset, select the file or files to
import and specify batch ID.
User Guide to SIMCA
10
Step Objective How to do it
9. Specifying the workset. The Workset dialog opens at Finish. Note
that the Create a batch level model
check box at the top of the Select Data
page is selected. Select the batch level
and batch conditions datasets to use and
then continue to the variables and
observations pages to include or exclude
etc.
10. Fitting the model. In the Fit model group click Autofit.
11. Reviewing the batch level
results and
performing diagnostics.
Create plots found on the Home tab to
detect trends, groupings, deviations, etc.
12. Using the model for
predictions.
On the Predict tab, use the predictionset
to see how new data fit with the active
model. The batch evolution predictionset is
automatically rearranged to a batch level
predictionset.
Batch project overview
Batch evolution
With batch data, you start by importing the batch evolution data and create the batch
evolution model, BEM.
Batch and phase identifiers and maturity variable
In the data, you must have a Batch ID (identifier), indicating the start and end of the
batch, and if phases are present, also a Phase ID. You may also have a variable
indicating the evolution of the batch or phase and its end point. This variable can be
Time or Maturity. You can have different Maturity variables for different phases.
Default unfitted models
Unfitted batch evolution models are implicitly created by SIMCA. When batches have
phases, each phase is fitted as one PLS class model with Time or Maturity as Y for
each phase. By default all variables in a phase are centered and scaled to unit variance.
Fitting the model
Click Autofit with the BEM marked to fit all models at once.
Reviewing the results
Display the results of the analysis of the workset batches in the Analysis Control
Charts group on the Batch tab, either as scores, DModX, Hotellings T2Range,
predicted time or maturity, or as individual variables.
Predictions
More datasets can be imported with new batches. The predictions can also be displayed
in Batch Control Charts in the same way by selecting the plot from the Prediction
Control Charts group on the Batch tab.
How to get started
11
Batch level
The batch level dataset is based on scores, original variables, or statistics of the
original variables for completed batches, obtained from the BEM. The model created
from these batch level datasets is named batch level model, BLM.
Creating batch level dataset
Create the batch level datasets by clicking Create Batch Level, in the Dataset group,
on the Batch tab. Specify the properties of the batch level datasets to create.
Creating batch level model
Clicking Finish in the Create Batch Level Dataset wizard opens the Workset dialog
with the Create a batch level model check box selected. Here you can select any
batch conditions and batch level dataset created from the same BEM. To import batch
conditions, click File | Import Dataset.
You can change the default model type to any desired model type allowed by the
workset specification. Alternatively specify hierarchical (Create Hierarchical Batch
Models on the Batch tab) or partial (select the Create partial models for each phase
check box in the Workset dialog) models.
Reviewing the results
Display the results of the analysis of the workset batches in the Diagnostics &
Interpretation group, on the Home tab.
Predictions
The batch evolution predictionset is automatically rearranged to a batch level
predictionset. This means that to make predictions for a BLM you select data from
batch evolution datasets and SIMCA takes care of the rest.
Project window for batch projects
The project window displays, for every model, one line summarizing the model results.
When batches have phases the PLS batch class models (one for every phase) are
grouped under an umbrella BEM. Both the BEM and BLM are grouped under an
umbrella called BMxx, xx is a sequential number.

Use the tutorial to get a guided tour for
building PC and PLS models.


Find the tutorial in Downloads on www.umetrics.com.

13
Introduction to multivariate
data analysis
Introduction
This chapter gives an introduction to multivariate data analysis of regular and batch
processes.
Content
- Process data analysis
- Batch data analysis
- Traditional data analysis: One or two variables at a time
- Multivariate approach
- Conclusion
Process data analysis
Any investigation of a real process or system is based on measurements (data). Fifty
years ago, measurement devices were expensive and few, and consequently the amount
of data, measured on processes, was limited; a temperature and a pressure here, a flow
rate there. The monitoring, display, and analysis of these few data was relatively
simple, and a few running charts of the data provided all available information about
the state of the process.
Today, batteries of sensors and on-line instruments are providing data from all parts of
the process in various forms, often at very short intervals. The masses of data are fed
into computers, re-computed into moving averages (per minute, hourly, daily, weekly,
etc.) and displayed and stored.
This change from a situation with few, fairly infrequent measurements, to many,
almost continuously measured variables has still not affected the manner in which
process data are treated, potentially leading to large losses of information. With
appropriate multivariate analytical methods, such as Principal Components (PC) and
Projection to Latent Structures (PLS) modeling in the SIMCA package, the masses of
process data can provide easy to grasp graphical information about the state of the
process, and relations between important sets of process variables. These multivariate
methods make efficient use of all pertinent data, with little loss of information.
User Guide to SIMCA
14
Process data properties
Before going into the analysis of process data, it may be useful to discuss their nature.
The data are usually measured at regular intervals over time; say every day, every
hour, or every minute. These intervals are often different for different variables. We
can recognize five categories, types, of process data: controlled process variables,
result variables, characteristics of raw material, intermediate result variables, and
uncontrolled variables. The variable types are described in the text that follows.
Controlled process variables
Related to the controlled setting of the conditions of the process, these are variables
that, in principle, can be changed, thereby affecting the results, the output, of the
process. We shall denote the values of these variables by x
ik
(observation i, variable k).
Examples of such variables are:
- T1, T2: the measured temperatures in reactors 1 and 2.
- P1, P2: the pressures in reactors 1 and 2.
- f12: the flow rate from reactor 1 to reactor 2.
Result variables
The availability of multivariate result variables (output, responses) measuring
important properties of the products coming out of the process dramatically increases
the possibilities to better understand and optimize a process. The result variables, are
denoted y
im
(data point i, y-variable m).
Examples of such variables are:
- y
i1
= yield of main product (%).
- y
i2
= impurity level (%).
- y
i3
= amount of side product no. 1 (%).
- y
i4
= tensile strength of product.
Characteristics of raw material
The characteristics of the raw material (inputs) are also denoted by x
ik
. These variables
are often of great importance for the process and the product properties, but usually
difficult or impossible to control.
Examples of such variables are:
- Concentrations of iron, coal, nickel and vanadium in feed.
- Fiber length distribution in input pulp.
- Gas chromatographic and spectroscopic (NMR, IR, etc.) analysis of feed
stock.
Intermediate result variables
Intermediate result variables are denoted by x
ik
, y
im
, or z
it
.
Examples of such variables are:
- Viscosity of output from reactor 1 (=input to reactor 2).
- Concentration of oxygen in output from reactor 1.
Introduction to multivariate data analysis
15
Uncontrolled variables
Uncontrolled variables are denoted by x
ik
, or z
it
.
Examples of such variables are:
- Air humidity, temperature of cooling water.
- Amount of oxygen in sparging nitrogen.
The purposes of measuring the data on the process and its input and environment are
to:
- Provide information allowing a better understanding of the process,
relationships between different parts of the process, which chemical or other
reactions occur inside reactors, etc.
- Yield information about the "state" of the process, recognizing trends,
peculiarities, etc., all to keep the process under proper control.
- Discover how the output is affected by the process and input variables in
order to improve product quality and minimize manufacturing costs,
pollution, etc.
Batch data analysis
With batch processes, K variables are measured on N batches at regular time intervals.
This gives a J x K matrix for each batch (J time points (X) times K variables).
Consequently, a set of N normal batches gives a three-way matrix of dimension (N x J
x K).
Variables
Batches
Time
One batch


Figure 1: Three-way table of historical data of, for example, N batches with J time
points, and K variables. Often the batches have different length (not shown in the
picture) and may also be divided in phases.
User Guide to SIMCA
16
Batch data properties
One of the differences between batch data and process data is that the batch process
has a finite duration. Each batch can also go through several phases before completion.
Since the batch data are collected batch wise they need reorganization before
importing.
Reorganization of 3D table before import
To be able to import the dataset, the 3-way data table has to be unfolded to preserve the
direction of the K variables (see figure 2). The resulting matrix has n
obs
= N x J
observations (rows) and K columns. Hence this matrix, X, has the individual
observation as a unit, and not the whole batches.
B1
B2
X Y
Time
B3

Bn
Variables
Batches
Recorded variables:
Ethanol,
Temperature,
Molasses flow,
NH3 feed,
Air flow,
Tank level,
pH.

Figure 2. The three-way matrix of figure 1 unfolded along the batch direction to give a
two-way matrix with N x J rows and K columns. Each row has the data (x
ijk
) from a
single batch observation (batch i, time j, variable k).
Introduction to multivariate data analysis
17
Traditional data analysis: One or two variables at a
time
To gain insight into the state of a process it is common to display important variables
and their change over time (figure 1). This works fairly well with up to 5 to 10
variables, but thereafter becomes increasingly difficult to comprehend. Furthermore,
these "time traces" reveal little about the relationships between different variables.


Figure 1. Two of the variables plotted against time on horizontal axis.
Scatter plots of pairs of variables are common complements to time traces (figure 2).
One hopes to identify correlations, thereby identifying important process variables
causing changes in the output variables.
The basic problem with pair-wise scatter plots is that they provide little information
about the real relationships between, on the one hand, input and process variables, and,
on the other hand, output variables. This is because the output (y) is influenced by
combinations of the input and process variables (x).
User Guide to SIMCA
18

Figure 2. Scatter plot of y8* (impurity) against x2in (one of the inputs).
Multivariate approach
With multivariate methods, one can investigate the relations between all variables in a
single context. These relationships can be displayed in plots as easy to understand as
time traces and pair-wise scatter plots.
The methods for multivariate process data analysis included in SIMCA are PC analysis
and modeling, PLS modeling, OPLS modeling, O2PLS modeling and more. The
principles and mathematics of the fit methods are briefly described later in the
introduction, and more thoroughly in the statistical appendix. References are also given
to additional pertinent literature.
Here we will indicate how to use these methods to solve typical problems in process
data analysis.
Summarizing a set of process variables
Data measured on a process are usually stored in some kind of database. A process
database containing the values of K variables for N data points can be regarded as a
table, a matrix. We denote the whole table by X. Each column in the table corresponds
to one variable (x
k
), and one row (x
i
) corresponds to the values observed at one point in
time.
Inputs or outputs or both can be displayed in multivariate control charts.
Summarizing batch data
Data measured batch wise are also often stored in some kind of database. One of the
differences between batch data and process data is that the batch process has a finite
duration. Each batch can also go through several phases before completion.
Inputs or outputs can be displayed in batch control charts.
Introduction to multivariate data analysis
19
PCA - Principal Component Analysis
Principal component analysis of a data table gives vectors of scores, with values t
ia
,
which summarize all the variables entering the analysis. It is customary to calculate
two or three score vectors, and then plot them against each other (tt-plots). This gives a
picture that is the best summary of the process behavior over time! In this plot we can
see trends, unusual behavior and other things of interest. With experience, one will be
able to recognize an area in this PC score plot in which the process remains under
"normal" operation, thus providing a multivariate control chart.
The score plot in combination with the loading plot, indicate the responsible variables
for deviations from normal operation.
The PCA also gives residuals, deviations between the data and the PC model, named
DModX. When these residuals are large, this indicates an abnormal behavior in the
process. To see this, we make a plot of the residual standard deviation, DModX
(residual distance, root mean square).
Observations with a DModX larger than the Dcrit are outliers. When DModX is twice
Dcrit they are strong outliers. This indicates that these observations are different from
the normal observations with respect to the correlation structure of the variables.
PLS - Partial Least Squares Projections to Latent
Structures
Relating the result variables to input and process variables
A common problem is process diagnostics. That means identifying those input and
process variables, X, that are responsible for the change in output variables, result
variables, Y. For this purpose, one often attempts to use multiple regression, which,
however, leads to great difficulties because process data usually do not possess the
correct properties for regression modeling. In particular, regression deals with each
result variable (y
m
) separately, and one therefore ends up with a set of models, one for
each output of interest. This makes interpretation and optimization difficult or
impossible.
To allow strict interpretation of "cause and effect", the data should be collected in a
careful experimentation using statistical design (with software such as MODDE). To
search for relationships between input and output in process logs is risky and often less
successful. This is because a process does not provide data with good information
content when the important factors are well controlled within small "control intervals".
PLS - Scores
PLS modeling has been developed explicitly for this type of situation with numerous,
often-correlated input and process variables and several to many result variables. One
just specifies which variables in the database that are predictors (X), and which
variables are dependent (Y), and PLS finds the relation between the two groups of
variables.
The PLS model is expressed as a set of X-score vectors, Y-score vectors, X-weight and
Y-weight vectors, for a set of PLS model dimensions. Each dimension (index a)
expresses a linear relation between an X-score vector (t
a
) and Y-score vector (u
a
). The
weight vectors of each model dimension express how the X-variables are combined to
form t
a
, and the Y-variables are combined to form u
a
. In this way the data are modeled
as a set of "factors" in X and Y and their relationships. Plots of the scores and weights
facilitate the model interpretation.
User Guide to SIMCA
20
PLS - Loadings
The PLS analysis results in model coefficients for the variables, called PLS-weights or
loadings. The weights for the X-variables, denoted w, indicate the importance of these
variables, how much they in a relative sense participate in the modeling of Y. The
weights for the Y-variables, denoted by c, indicate which Y-variables are modeled in
the respective PLS model dimensions.
When these coefficients are plotted in a w*c plot, we obtain a picture showing the
relationships between X and Y, those X-variables that are important, which Y-
variables are related to which X, etc.
PLS - Residuals
Analogous to PC, PLS gives residuals, both on the Y-side and on the X-side. The
standard deviations of these (residual distances) can be plotted just as for PCA to give
a third "SPC" plot (Statistical Process Control) showing if the process is behaving
normally or not.
OPLS and O2PLS - Orthogonal Partial Least Squares
OPLS is an extension of PLS and addresses the regression problem. OPLS separates
the systematic variation in X into two parts, one part that is correlated (predictive) to Y
and one part that is uncorrelated (orthogonal) to Y. This gives improved model
interpretability. In the single-Y case, there is only one predictive component, and all
components beyond the first one reflect orthogonal variation. However, with multiple
Y-variables there can be more than one predictive OPLS component.
O2PLS is yet another extension of PLS and addresses the data integration problem.
Thus, in the two-block (X/Y) context, O2PLS examines which information overlaps
between the two data tables and which information is unique to a specific data table (X
or Y). O2PLS accomplishes this task by a flexible model structure incorporating three
types of components, that is,
(i) components expressing the joint X/Y information overlap,
(ii) components expressing what is unique to X, and
(iii) components expressing what is unique to Y.
For both OPLS and O2PLS the different components are interpretable the usual way,
since the scores, loadings, and residual-based parameters with a familiar meaning are
preserved.
Batch modeling in SIMCA
A SIMCA batch project consists of one or more batch evolution models, BEM, and
optionally one or more batch level models, BLM:
1. The BEM has several observations per batch with the variables measured
during the evolution of the batch. Batch evolution models are commonly
applied for real time process monitoring.
2. The BLM consisting of the completed batches, with one batch being one
observation (matrix row). The variables of the BLM are the scores, original
variables, or summary variables (statistics of the original variables) of the
BEM at every time point folded out side-wise. Batch level models are
commonly applied to investigate variations between batches, sites,
campaigns or to study impact of process variations on product quality.
Introduction to multivariate data analysis
21
A BM, sequentially numbered, holds the BEM and the BLM-models built on datasets
created from that BEM.
Batches may be divided into phases.
Batch evolution modeling
When importing the dataset in SIMCA, an index y-variable starting at 0 is constructed
if no maturity is specified. This is the "relative local batch time".
The default model, after completed import in SIMCA, is a PLS model for the centered
and scaled X and y, or "relative local batch time" as y if no other y was specified. If a
maturity variable, recorded relative to the start of each batch, was defined as y, it is the
variable used as y.
When fitting the batch evolution PLS model, Autofit will stop when 85% of X has
been explained, but will also give at least 3 components, as long as the number of
components does not exceed number of variables/2. This is necessary in order to
ensure that much of the X-matrix is "explained". This results in an (n
obs
x A) score
matrix, T, plus PLS weight and loading matrices W and P (of dimensions K x A).
Modeling the evolution of batches in the workset
At the start of a batch project, when phases are present, the BEM workset is created
with an unfitted model for every phase. The BEM is the wrapper grouping all the PLS
class models (one for every phase). Maturity or time, as specified for each phase, is
used as Y, the other variables as X, and all are centered and scaled to Unit Variance.
When batches have no phases, there is only one model in the BEM, with time or
maturity as Y and all other variables as X, centered and scaled to Unit Variance.

Note: A model is positioned in a BEM only when the data was imported as
batch and it is a PLS model with Time or Maturity specified as y-variable.
To fit the model(s), mark the BEM and click Autofit.

User Guide to SIMCA
22
The BEM is built on the good batches when the goal is to define a process monitoring
model with the ideal evolutionary trace. Around this trace, 3 SD limits are used to
define the acceptable variation.
Examine the batch evolution models using the Batch control charts (BCC), starting
with the Variable BCC to ensure that the process data is good and reliable for all
batches. Look for outlier batches in batch control chart scores and/or the DModX plots.
Interpret deviating batches with contribution plots.
Batches with different length
Alignment of data is done differently depending on configuration of the time/ maturity
variables. For batches of similar length the impact is small. The default y configuration
aligns the batch length by cutting the longer batches at median (or average) batch
length. Batches that are shorter than the median (or average) batch, are filled up with
the last value up to the median batch length.
The used length in average batch is the shorter of the median batch length and the
average batch length rounded down.
Batches with several phases
When batches have several phases, observations are organized in classes, one class for
every phase. Separate PLS class models are fitted for every phase.
Smoothing the maturity variable
If a maturity variable is used as the response variable Y, indicating the degree of
completion of a phase, the maturity variable has to be strictly monotonic in a given
phase, but can be ascending or descending.
Different phases can have different maturity variables. Different phases can have
ascending or descending maturity.
When maturity variables are not strictly monotonic, they are smoothed within each
phase. The smoothing is done by fitting a constrained quadratic polynomial, using a
piece wise least squares algorithm. This is to ensure that the maturity is strictly
monotonic within a phase. The smoothed maturity is used by default, both as the
response variable Y, and for alignment.
Result variables
The score matrix T is a good summary of X, and PLS focuses this summary on y (local
time). Hence the first column of t (t
1
) will contain strong contributions of those X-
variables that vary monotonously with y, i.e., either increase or decrease with time.
The second component (t
2
) is an aggregate mainly of those variables that change
quadratically with time, i.e., first go up to a maximum, and then decrease, or first go
down, and then up. The third component catches the variables having a cubic behavior,
etc.
The value of "local time" predicted by the PLS model (with the number of components
determined by cross validation ), y
pred
, is very suitable as a "maturity index" that can be
used to indicate how far the batch has evolved, if it is ready for termination, etc.
Introduction to multivariate data analysis
23
Calculating limits to build control charts
To monitor the evolution of new batches, one calculates the typical evolution trace of a
normal batch in terms of the scores, DModX, etc. This is in short done as follows:
1. Reorganization: First the scores of the BEM are chopped up and
reorganized so that the scores of one batch form one row vector (t
1
followed
by t
2
, followed by t
3
, etc.) in a matrix S
T
. This matrix has N rows (one per
batch) and A x J (AJ) columns from the A score vectors and the J "time
points" per batch. When batches have phases, the scores of each batch are
collected from the different PLS phase models and the calculations are done
separately for the different phases.
t1
Batch
Average
StDev
Avg + 3 sd
Avg 3 sd
Time

Figure 4. The score values for each batch are arranged as row
vectors under each other, giving an N x (J x A) matrix, S
T
. Above we
see only the first part of S
T
corresponding to the 1st component (t
1
).
From this matrix one calculates the averages and standard deviations
(SDs) of the matrix columns, and then control intervals as the
averages 3 SD.
2. Calculating tolerance limits for score batch control charts: Now the
matrix S
T
is used to derive minimum and maximum tolerated values at each
time point (j) of t
j1
, t
j2
, t
j3
, t
jA
from the column averages and standard
deviations of S
T
, as indicated in figure 4. In MSPC one would typically use
tolerance intervals of the average 3 SD at each time point. Hence, we now
have for each score vector an average trace with upper and lower tolerance
limits (figure 4). This specifies the normal evolution traces, one for each
score component, of new batches.
3. Calculating tolerance limits for y predicted batch control chart: An
additional trace is constructed from the values of predicted y (local time or
maturity), for each time point of each batch. The predicted y-values should
be fairly close to the "real" y-value for a batch evolving at the "normal" rate.
SIMCA forms a matrix R
Y
with the traces of predicted y of each batch as
rows, and then analyzing it in the same way as S
T
above, and gives tolerance
intervals of y
pred
for each time point.
4. Calculating tolerance limits for DModX batch control chart: An
additional trace is constructed from the values of the residual standard
deviation (distance to the model, DModX), for each time point of each batch.
The residual standard deviation (DModX) is calculated from the
residuals, i.e., after subtracting t
i
x P' from the scaled and
centered observation x-vector. For an observation to be judged OK (non-
deviating), DModX should be smaller than a critical limit, D
crit
calculated
from the F-distribution.
Note: When batches have phases, the limit calculations are done by phase.
User Guide to SIMCA
24
Monitoring the evolution of new batches
The batch evolution data of new batches are inserted into the PLS model, giving
predicted values of the score values t
1
to t
A
. These results can now be plotted in the
appropriate batch control charts, with limits derived as described in Calculating limits
to build control charts. These charts indicate whether the batches are starting
normally or not. If the values are outside the normal ranges, contribution plots based
on the x-values or the residuals indicate which variables together are related to the
deviations.
In addition, the PLS model gives predicted values of y (local time or maturity index).
Plotting y
pred
vs. y
obs
gives a very interesting indication of whether the batches are
developing too quickly (over-mature, y
pred
> y
obs
) or too slowly (under-mature, y
pred
<
y
obs
).
When batches have phases, the appropriate PLS phase model is used to predict the t
scores of every phase. The scores are collected for every batch and results are plotted
in the control charts.
Batch level modeling
This section describes modeling of entire batches as observations (rows) using batch
level modeling.
Why batch level modeling?
Batch level modeling may be advantageous in the case that the measurement of Y is
laborious and time consuming, and one wishes to make an initial fast quality
assessment before the real Y is available. This is often the case in biotechnological
production, where extensive analysis and testing of the final products is made before
acceptance (or rejection).
The batch level model (BLM) also allows the full interpretation of the batch data.
Groups of batches may be discovered, outliers will be found, etc. Critical time periods
will be indicated by "periods" of large weights, loadings, and VIP values. Analogously,
important factors in the batch conditions will stand out.
Creating the batch level model
The batch level model can be created based on scores, original variables, summary
variables (raw data statistics), duration and endpoint, plus batch conditions (Z).
Introduction to multivariate data analysis
25
Modeling batch level
The objective of batch level modeling is to make a model of the whole batch in order
to understand how Y is influenced by the combination of batch conditions and the
batch evolution. This model will be based on (when available) the batch conditions (Z),
the evolution trace matrix S
T
, and when applicable the properties and quality of the
completed batch (Y), as shown in figure 5. The resulting model is used to predict Y for
new batches from their batch conditions and their evolution traces.

Figure 5: The data for batch level modeling. Each row has the data from one batch.
The left matrix contains data describing initial batch conditions. The middle matrix
contains the unfolded scores/raw variables, which describe the evolution of each
batch. The optional right matrix (Y) contains the responses, the properties of the
complete batch such as yield, purity, activity, etc.
In the case that Y-values do not exist, the X-matrix can still be used to develop a
PC model. We see that this is the same type of model as the X-part of PLS in the batch
evolution model. The difference is that in PCA the scores (T) are calculated to give an
optimal summary of X, while in PLS this optimality is somewhat relaxed to make T
scores better predictors of Y.
This PC model can then be used to classify new batches as normal (similar = well
fitting to the PC model) or non-normal (far from the model).
Predicting results or classifying whole new batches
The batch level model is used to predict quality of the final batch (Y), or classify the
batch as normal good batch or bad batch or predict their final quality.
If you have fitted hierarchical models of the batch level data at different percentage of
completion, you can classify or predict the final quality of new batches as they are
evolving when they reach the percent of completion of the models.
If the values are outside the limits, contribution plots indicate which variables together
are related to the deviations.
Create a batch level model
When creating a batch level dataset, clicking Finish opens the Workset dialog with
the Create a batch level model check box selected.
Select the datasets to use and then use the other pages in the dialog for editing the
workset as desired. Clicking OK in the Workset dialog creates the unfitted BLM.
Note: Selecting and clearing dataset check boxes in the Select Data page
result in resetting the workset. All variable and observation/batch settings are
cleared along with any changes in transformations, lags, expansions, scaling
and trimming and the result is the default workset.
User Guide to SIMCA
26
Hierarchical batch models
For batch level datasets you can fit hierarchical models of the whole batch, for
different levels of completion:
- One model for each phase and component in BEM.
- One model for each phase.
- Sequential models covering parts of the batch completion, for example for
25%, 50%, 75% and 100% completion for models without phases.
These models can predict the final batch quality or classify the batch, before the batch
is finished.
Conclusion
Projection methods such as PCA, PLS, OPLS and O2PLS find the information in
masses of process data by projecting these data down on a few "scores". These scores
provide a very good summary of process data tables (X and Y), and score plots display
this in an easy to grasp form. The coefficients of the projections, i.e., how the variables
are combined to form the scores, are called loadings or weights. Their plots show the
importance of the variables, their similarity, their connection, and other things of
interest.
The parts of the data not seen in the score plots, i.e., the residuals, are displayed (in
summarized form) in the DModX and DModY plots (row residual standard
deviations).
For further reading, see the Multi- and Megavariate Data Analysis book and the
reference list.

27
Overview of SIMCA
Overview of SIMCA
This chapter describes the following:
- Application icon and symbol.
- SIMCA Projects.
- SIMCA window interface.
- Ribbon content.
- Shortcut menu.
- Conventions including limitations and missing values representation.
- Presentation of SIMCA results
Application icon and symbol

The application icon is a purple circle with a white S inside (below).

Projects
Multivariate data analysis in SIMCA is organized into projects. You can think of a
project as a file folder containing all the information related to the analysis of a number
of datasets. Projects are default named after the first selected dataset in SIMCA import.
This information is organized in the following components:
- Datasets used for the analysis.
- Other datasets used for predictions and model validation.
- Workset and multivariate models.
User Guide to SIMCA
28
Datasets
For continuous processes, the data are observations made sequentially in time, usually
at a constant sampling interval. Such data are called time series. In the analysis of time
series, the observed outcome at time t is often modeled as a function of the previous
observations at time t
1
, t
2,
etc.
With PLS, the time series analysis is performed by creating lagged variables, i.e.,
variables shifted in time, and fitting the model. For PLS time series analysis to give
correct results the observations must be contiguous in time (the sampling interval is
constant), and with no missing observations.
A batch process has a finite duration. Each batch can go through several phases before
completion. Since the batch data are collected batch wise they need reorganization
before importing.
Workset
A workset is a subset or all of the data in the selected datasets with a particular
treatment of the variables, i.e., role (predictor variables X, or responses Y),
scaling, transformation, lagging, etc.
The workset is by default the first dataset with all imported variables and observations
included and the variables centered and scaled to unit variance, for regular projects.
For batch projects the default workset is all batch evolution datasets included when
creating the project.
When you define a workset, SIMCA creates an unfitted model, and both the model and
the workset are identified by the same name.
Models
Models are mathematical representations of your process and are developed using the
data specified in the workset and with a specified model type, for example, PCA-X,
PCA-Y, PCA-X&Y, PLS, etc.
With a specified workset, you may be able to fit several models by selecting different
model types. For example, if you have specified a workset by defining the X-variables
(predictors) and the Y-variables (responses), you may first fit a PC of the responses (Y-
block) and then fit a PLS model, by just selecting another model type.
Models are created by SIMCA when:
1. You specify a workset, New/New as Model.
2. You make a selection under Change Model Type with a fitted model,
You work with one model at a time, the active model. This model is marked in the
Project Window and displayed in the caption of the Project Window while the
project window is docked.
There is always an active model. You may have several models open, but only one of
them is active.
Overview of SIMCA
29
SIMCA window
The SIMCA window consists of the ribbon with regular and context tabs, the quick
access toolbar, and dockable windows.

User Guide to SIMCA
30
SIMCA ribbon description

Description of the ribbon using the terminology according to Microsoft:
- The File tab, which presents a menu of commands that involve doing
something to or with a document or workspace, such as file-related
commands.
- A Quick Access Toolbar, which is a customizable toolbar that displays
frequently used commands.
- Core tabs are the tabs that are always displayed.
- Contextual tabs, which are displayed only when a plot or list is open, Tools
and Layout or when items are marked in a plot or list, Marked Items.
- Galleries, which are lists of commands or options presented graphically. A
results-based gallery illustrates the effect of the commands or options instead
of the commands themselves.
- Enhanced tooltips, which concisely explain their associated commands and
give the shortcut keys. Enhanced tooltips reduce the need for command-
related Help.
- Dialog box launchers, which are buttons at the bottom of some groups that
open dialog boxes containing features related to the group
License information
The Microsoft Office User Interface is subject to protection under U.S. and
international intellectual property laws and is used by MKS Umetrics under license
from Microsoft.
Overview of SIMCA
31
Ribbon
Available functionality can be accessed through the SIMCA interface.

The SIMCA interface includes:
- the Quick Access Toolbar
- the regular tabs File, Home, Data, Batch, Analyze, Predict, Plot/List, and
View
- the context tabs Tools, Layout, and Marked Items
- and the Help menu.
On the regular and context tabs the commands are parted in groups. Some of the
groups have a dialog dispatch launcher which opens a dialog closely connected to the
features in the group.
For a short introduction see the sections that follow here. For more, see the respective
chapters.
Quick Access Toolbar
The Quick Access Toolbar allows you to have your favorite commands easily
accessible. By default it has:
- The standard commands Save, Paste, Cut, Copy, Undo and Select All.
- Find Allows you to search in plots.
File tab
The following is available on the File tab:
- New Regular/Batch Project Creates new SIMCA projects after selecting
a dataset in SIMCA import.
- Open Opens old project.
- Save Saves the current project.
- Save As | Save Project As/Save Plot As/Save List As Saves the current
project/plot/list to the specified path and name.
- Print | Print/Print Preview/Print Setup.
- Manage:
Encrypt Encrypts the project using the user entered password.
Compact Project File Compacts the project file by removing selected
parts from the .usp.
Delete Project - deletes the current project.
- Import Dataset Imports selected datasets.
User Guide to SIMCA
32
- Generate report - Creates a HTML report with the selected template or an
existing report.
- Close Closes the current project.
- Recent files - Lists the projects most recently opened.
- Project Options button - Displays the current default options for the current
project.
- SIMCA Options button - Displays the current default options for SIMCA.
- Exit - Closes the current instance of SIMCA.
Home
On the Home tab you can change the model type, fit the model and review the fit. Plots
and lists to analyze the model are available from this tab.
On the Home tab the following is available:
Dataset group
- Dataset Lists and opens all available datasets.
Workset group
The workset is a working copy of the selected datasets with certain properties defined.
Note: The datasets are NOT affected when you exclude variables/observations
or by any other operation in the workset.
In the Workset group the following is available:
- Statistics | Descriptive Statistics Displays statistics of selected variables
of the active model.
- Statistics | Correlation Matrix Displays the correlations between the
variables for the active model.
- New Creates the default workset consisting of the first listed dataset with
the variables set as X or Y as specified at import, and scaled according to the
specified scaling in Project Options. For batch all datasets imported when
creating the project are included in the default workset.
- New As Creates a new workset as a copy of the selected model.
- Edit Opens the selected specified model for editing. If the model has
already been fitted that model will be deleted and at OK a new unfitted
model is created.
- Delete Deletes the selected workset, and consequently also the model.
- Change Model Type - Displays the available model types:
- Overview - PCA-X, PCA-Y, PCA-X&Y
- Regression - PLS, OPLS, O2PLS
- Discriminant analysis - PLS-DA, OPLS-DA, O2PLS-DA
- Class - PCA, PLS, OPLS, O2PLS
- Clustering - PLS-Tree
Overview of SIMCA
33
- Model Options Available from the dialog box launcher (arrow beneath the
Change Model Type arrow) and displays the options of the active model.
Fit model group
For the model or model group (CM/BEM):
- Autofit Autofits according to the autofitting rules.
- Two First Calculates the two first components.
- Add Calculates the next component.
- Remove Removes the last component.
Diagnostics & Interpretation group
Overview - Displays the 4 default plots X/Y Overview, Scores, Loadings,
DModX.
Summary of Fit:
- Summary of Fit plot and list Displays the cumulative fit over all variables
for each component.
- X/Y Overview plot and list Displays the cumulative fit of all variables for
each x-variable for PCA and each y-variable for PLS, OPLS and O2PLS.
- X/Y Component Displays the fit of a variable for each component.
- OPLS/O2PLS Overview - Displays OPLS and O2PLS specific R2.
- Component Contribution Displays the contribution of a model
component to the fit.
Review the fit and investigate diagnostics:
- Scores t1 vs. t2, Num vs. t1, etc. Displays trends, groupings, outliers.
- Loadings p1 vs. p2, w*c1vs. w*c2, etc. Displays important variables,
variable correlations.
- Hotellings T2 Displays a measure of how far each observation is from the
center.
- DModX for X or Y Displays the distance from the observation to the
model, by default after the last component.
- Observed vs. Predicted Displays the actual value versus that observation
predicted by the model for the selected response (y) variable.
- Coefficients Displays the coefficient of the model for each response (y)
variable cumulatively over all components.
- VIP Displays the overall importance of each variable (x) on all responses
(Y) cumulatively over all components.
Data
On the Data tab the following is available:
Modify Dataset group
- Merge Merges the selected datasets and deletes the second dataset
(source).
User Guide to SIMCA
34
- Split - Splits the selected dataset as specified in the dialog. Only available
for independent datasets.
- Transpose Transposes the selected dataset and deletes dependent models
and datasets. Not available for batch projects or dependent datasets.
- Delete Dataset - Deletes the selected dataset and any dependent models and
datasets.
- Generate variables Creates new variables as functions of the existing ones
or from model results.
- Local Centering - Import local centering or View already imported local
centering data.
Filters group
- Spectral filters - Filters the selected dataset using the selected filter creating
a dataset in the current project. The available filters are: Derivative (1st, 2nd,
and 3rd derivative), Multiplicative Signal Correction (MSC), Standard
Normal Variate (SNV), Row Center, Savitzky-Golay, EWMA,
Wavelet Compression, Wavelet Denoising, Orthogonal signal correction
(OSC) and chained filters.
- Time Series Filters - Time series filters the selected dataset using Wavelet
Compress Time Series and Wavelet Denoising/Decimation creating a new
dataset in the current project.
Summary group
- Dataset Summary Displays a summary of the performed filtering.
Available for filtered datasets.
- Missing Value Map - Displays an overview of a dataset with respect to
missing values.
- Trimming Overview Opens an overview window of the trimming. Only
available after trimming.
- Spectra - Displays ObsDS, for the variables specified as X, of the selected
dataset.
Base Model group
- Hierarchical - Specifies the current model as hierarchical base model
creating a hierarchical dataset.
- Non Hierarchical - Removes the hierarchical base dataset.
Batch
Analysis Control Charts and Prediction Control Charts groups
- Batch Control Charts Displays scores, distance to model, variable,
Hotellings T2, or predicted y with limits in plots and lists for models and
predictions (BEM only).
Overview of SIMCA
35
Time/Maturity group
- Observed vs. Smoothed Y - Displays the maturity for selected batch in
original and treated form (BEM only).
- Unaligned vs. Aligned - Displays the selected vector for the selected batch
aligned and unaligned (BEM only).
Dataset group
- Create Batch Level Creates the batch level dataset according to the
selections made by the user. Available with a BEM marked only.
- Create Hierarchical Batch Models Creates hierarchical batch models
according the selections in the dialog. Only available for batch level
datasets.
Variable Summary group
- Variable Importance Plot Displays the overall importance of the variable
over the whole evolution on the final quality of the batch. Available for BLM
only.
Analyze
Analysis group
- Biplot Displays scores and loadings superimposed.
- Inner Relation - Displays t1 vs u1, t2 vs u2 etc.
- S-plots - S-plot, S-line, SUS-Plot display p in forms informative for OPLS
with one Y.
- Contribution Displays variables likely to be causes for upsets in the score,
DModX, Hotellings T2Range, or predicted response plots.
- RMSECV - Indicates predictive power.
- Y-Related Profiles - Coefficients rotated displaying the pure profiles of the
underlying constituents in X using the assumption of additive Y-variables for
OPLS and O2PLS.
- Residuals N-Plot Displays the residuals on the normal probability chart for
the selected Y.
Validate group
- Permutations Displays the validity and the degree of over fit for the
model (only PLS).
- CV-ANOVA For assessing the reliability of the model.
- CV Scores - Displays the cross-validated complement to the regular
scores plot
Clustering group
- HCA Displays the clusters from the cluster analysis in a tree diagram.
- PLS-Tree - Displays the cluster analysis result using PLS-Trees.
User Guide to SIMCA
36
Predict
On the Predict tab you can select a predictionset, display the predictionset as a
spreadsheet, and display the prediction results using the active model.
The following is available:
Specify Predictionset group
- Specify Defines the predictionset in the Specify Predictionset dialog.
- As Dataset - Predictionset is the selected dataset.
- As Workset - Predictionset is the current workset.
- Complement Workset/Complement WS Batches - Predictionset is the
observations/batches not included in the current model.
- Class - Predictionset is the selected class.
- Delete Predictionset - Deletes the selected predictionset.
List group
- Prediction List Displays the predictionset, some predicted vectors such as
membership probability, scores, and the predicted response for PLS models.
Plots group
- Y PS Displays the predictions for the selected response (y).
- Scores PS Displays the predicted scores.
- Hotellings T2 PS Displays a measure of how far the predicted value for
the observation is from the center
- DModX PS for X or Y Displays the distance from the predicted value for
the observation to the model.
- Predicted Control Charts Displays the selected vector in the selected
control chart: Shewhart, CUSUM, EWMA, or EWMA/Shewhart.
- Contribution plots Show variables likely to be causes for upsets in the
predicted score, distance to model X, Hotellings T2Range, or predicted
response plots.
- Time Series - Displays the selected vector in a line plot.
- Coomans' plot Shows class separation for two selected models.
- Classification List - Displays the classification of a predictionset with
respect to all the class models.
- Misclassification Table Summarizes the number of observations, with
known class belonging, correctly classified in class or PLS-DA models.
Plot/List
The Plot/List tab holds general facilities for plotting and listing model and prediction
results.
Overview of SIMCA
37
The following types of plots are provided:
Standard Plots group
- Scatter Plots
- Scatter 3D Plots
- Line Plots
- Column Plots
- Lists
Control Charts group
- Time Series Plots
- Control Charts
Custom Plots group
- Histogram
- Normal Probability
- Response Contour
- Response Surface
- Wavelet Structure
- Wavelet Power Spectrum
- Step Response Plot
View
On the View tab the following is available:
Show/Hide group Shows or hides the following dockable windows:
- Audit Trail Displays all actions taken in the project.
- Analysis Advisor - Explains all plots on the Home tab.
- Favorites Holds a shortcut to selected plots, lists, and commands.
- Item Information Displays information about the marked items (points).
- Marked Items Displays the items, observations or variables, marked in a
plot.
- Model Window Displays the model window of the active model.
- Observations Displays the observations in the active model.
- Project Window - Displays the models of the current project in the Project
Window.
- Quick Info Displays information about the marked observation or variable.
- Variables Displays the variables in the active model.
- Status Bar The Status Bar displays an explanation when positioning the
cursor over a button.
User Guide to SIMCA
38
- Full Screen - Displays the ribbon when clicking a tab but not otherwise.
Add group
- Add to Favorites Adds the current plot or list to the Favorites window.
- Add to Report Adds the current plot or list to the HTML report supported
by SIMCA.
Window group
Use the Window group to access the standard window commands.
Tools contextual tab
The Tools tab becomes available when opening a plot or list. The content is context
sensitive. Changes introduced on the Tools tab apply to the active plot or list.
Change Type group
Change the plot type to
- Scatter
- Line
- Column
- List
For batches:
- Sources of variation - Converts the column plot to a sources of variation
plot. Only available with BLM including one or more raw or score-variables.
- Out of Control Summary - Displays the percentage outside the displayed
limits.
- Batch Control Chart - Available when the Out of Control Summary plot
(OOC plot) is displayed and displays the batches in the OOC plot.
For dendrograms
- Create List - Displays the background to the plot.
- Merge List - Displays the calculations behind the HCA dendrogram.
Data group
For the current plot or list the Data group:
- Model - Displays the active model.
- Comp - Displays the current component.
- Batch - Displays the selected batches.
- Variable - Displays the selected variable in for instance a BCC variable plot.
- Observation - Displays the selected observations in for instance a spectral
plot.
Overview of SIMCA
39
Properties group
The Properties group contains features also available in the Properties dialog for the
current plot or list. The Properties group displays the properties of the plot, list, or
window
- Color by - Applies the selected coloring.
- Labels - Displays the selected labels.
- Size by - Applies the selected sizing.
- Dialog box launcher - Opens the Properties dialog.
Highlight group
- Select Selection modes.
- Find - Opens the Find dialog.
- Series - Grays all but the series hovered.
- Sort - Sorts lists, spreadsheets and column plots as specified.
Zoom group
- Zoom Zooms according to the selected zooming type.
- Zoom Out - Zooms out one step.
Layout contextual tab
Template group
- Load - Switches to the selected template for all open and future plots.
- Save as - Saves the current plot formatting to a plot formatting file.
- Save as default - Saves the current plot formatting to the default plot
formatting.
- Open templates folder - Opens the folder where the custom templates are
stored.
- Restore Defaults - Restores the default template.
Show group
- Maximize Plot Area - Quick access to showing/hiding the header, legend,
footer etc.
- Header
- Footer
- Legend
- Axis Titles
- Axes
- Timestamp
- Regression Line - Displays the regression line and equation in the current
plot.
User Guide to SIMCA
40
- Resolve Coefficients - Available for hierarchical top models and results in
coefficients for base models.
Format Plot group
- Format Plot - Opens the Format Plot dialog.
Marked Items contextual tab
Create from Marked Items group
- Scatter - Creates a scatter plot holding only the marked items.
- Line - Creates a line plot holding only the marked items.
- Column - Creates a column plot holding only the marked items.
- List - Creates a list holding only the marked items.
Drill Down group
Contribution plots
- Point to Average Comparison
- Group to Average Comparison
- Point to Point Comparison
- Point to Group Comparison
- Group to Group Comparison
Line plots
- Plot XObs - Plots the marked observations X-part in a line plot using a
numerical ID as x-axis if available.
- Plot YObs - Plots the marked observations Y-part in a line plot using a
numerical ID as x-axis if available.
- Variable Trend Plot - Plots the marked variables in a line plot.
Modify Model group
- Exclude - Excludes the marked observations or variables.
- Create new BEM and BLM without marked batches - Available after
marking in a BLM score plot. Excludes the batches in the BEM, fits the
BEM and automatically creates the new BLM.
- Include - Adds the observations or variables to the unfitted model. Creates a
new model holding only the included observations or variables if there is no
unfitted model.
- Class - Depending on what is marked No Class, Class, Create Class Models,
Create OPLS-DA, and Create PLS-DA may be available enabling quick
creation of class/DA models.
Layout group
- Labels - Displays the selected label on the marked points.
- Hide - Hides the marked points.
Overview of SIMCA
41
- Show All - Shows all items.
- Format Symbol - Opens the Format Plot dialog with the new Custom node
enabling customizing the marked items style.
- Format Label - Opens the Format Plot dialog with the new Custom node
enabling customizing the marked items labels.
Help
Standard help menu providing access to the SIMCA help, the registration and About
SIMCA dialogs.
Shortcut menu
Right-clicking any plot or list in SIMCA opens the context sensitive shortcut menu that
includes general commands, such as Copy, Print, Format Plot, and Properties.
Conventions
Limitations in project names
The project name length (including path) cannot be larger than 255 characters.
Limitations in observation and variable identifiers
Observation and variable identifiers (names) can be up to 256 characters long.
The primary observation and variable identifiers must be unique. When importing
more datasets you are forced to have unique primary identifiers.
To be able to use a dataset as predictionset, the primary variable identifiers have to be
identical to those of the model.
Case sensitivity
SIMCA is case insensitive. Lower or upper case in names will be displayed as entered,
but for all comparisons lower or upper case are considered the same.
Menu and tab reference syntax
In this user guide we use the following syntax when referring to the File tab
commands:
- Click Tab | Command. An example: Click File | Save As | Save Project As.
- On the Tab tab, click Command. An example: On the File tab, click New
Regular Project.
- Click Command on the Tab tab. An example: Click New Regular Project
on the File tab.
In this user guide we use the following syntax when referring to tab
commands/buttons:
- On the Tab tab, in the Group group, click Command. An example: On the
Home tab, in the Workset group, click New.
User Guide to SIMCA
42
- Click Command, in the Group group, on the Tab tab. An example: Click
New in the Workset group on the Home tab.
- Click Tab | Command | Menu item. An example: Click Home | Statistics |
Descriptive Statistics.
In this user guide we use the following syntax when referring to a tab/menu item or
gallery item accessed by clicking a button:
- On the Tab tab, in the Group group, click Button | Menu item. An
example: On the Home tab, in the Diagnostics & Interpretation group,
click Summary of Fit | X/Y Overview.
- On the Tab tab, in the Group group, click Button, and then click Menu
Item/Gallery Item. An example for menu: On the Home tab, in the
Diagnostics & Interpretation group, click Summary of Fit, and then click
X/Y Overview. An example for gallery: On the Home tab, in the
Diagnostics & Interpretation group, click Scores, and then click Line.
- Click Tab | Button | Menu item. An example for menu: Click Home |
Summary of Fit | X/Y Overview. An example for gallery: Click Home |
Scores | Line.
Select and mark
Select or Mark an item in plots, lists, or menus signifies clicking the item leaving it
highlighted.
Vector and matrix representation
Capital letters signify matrices, for ex. X, Y, P, T.
Letters typed in lower case signify vectors, for ex. x, y, p, t.
Prime <> represents that the vector is a row vector. For example, t signifies a column
vector with the values found under each other while p signifies a row vector where the
values are found after each other.
Overview of SIMCA
43
Missing values representation
Missing values are represented by blank or space < >.
You can select another value to also represent missing in the SIMCA import Options
dialog, opened from the File tab in the SIMCA import, Missing Value field.
Note that 99 is assumed to be a valid value in your data and therefore automatically
changed it to 99.0001 to avoid confusion in the internal calculations in SIMCA.
If -99 represents missing value, enter -99.0001 in the Missing Value field in Options
in the SIMCA import.

Consequences of missing values
The present NIPALS algorithm works with missing data as long as they are relatively
randomly distributed, i.e., do not occur with a systematic pattern. Otherwise the results
of a model fitting can be very misleading. The default is to warn when the missing
value content in an observation or variable exceeds 50%.
User Guide to SIMCA
44
Presentation of SIMCA results
Plots for publications
To create and save a plot of user defined resolution suitable for publication, follow the
steps below:
1. Create desired plot in SIMCA and edit if necessary so that the plot has the
desired appearance and content.
2. File | Print; the Print dialog opens.
3. Under Printer, Name: select Microsoft XPS Document Writer, OK, enter
file name and save to preferred location.
4. Download and install Converter for XPS Documents from
http://www.openxml.biz/xpsconverter.html.
5. Start the program, File | Open and select *.xps file.
6. File | Convert to, select Image Type (e.g. TIFF is required by many
publishers) and Resolution (often a minimum of 300 dpi is required).
7. Choose folder to save results in and click OK.
Plots for reports
For suitable formats to copy or save, use one of the Umetrics predefined, see the Save
As subsection in the File chapter.

45
Quick Access Toolbar
Quick Access Toolbar
The Quick Access Toolbar is by default positioned at the top above the tabs and allows
you to have your favorite commands easily accessible. The Quick Access Toolbar is
standard in Windows software with ribbons.

The Quick Access Toolbar in SIMCA default displays the following features:
- Save - CTRL+S
- Paste - CTRL+V
- Cut - CTRL+X
- Copy - CTRL+C
- Undo - CTRL+Z
- Select All - CTRL+A
- Find CTRL+F.
Note: Undo is unavailable after adding or deleting rows/columns.
Customizing Quick Access Toolbar
You can add buttons to and remove buttons from the Quick Access Toolbar.
Removing buttons from Quick Access Toolbar
To remove a button from the Quick Access Toolbar:
- Right-click the button and click Remove from Quick Access Toolbar.
- Click the arrow to the right of the available Quick Access Toolbar buttons
and clear the relevant check box.

User Guide to SIMCA
46

Adding buttons to Quick Access Toolbar
To add a button to the Quick Access Toolbar, right-click the button on the ribbon and
click Add to Quick Access Toolbar.

Moving the Quick Access Toolbar
To position the Quick Access Toolbar closer to the workspace you can select to
position it below the ribbon by clicking the Show Below the Ribbon in the Quick
Access Toolbar menu. Click Show Above the Ribbon to move it back to its original
position.
Minimizing the ribbon
To maximize the workspace you can minimize the ribbon so that it folds away like a
menu.
The Minimize the Ribbon command is available:
- When you right-click a button in the Quick Access Toolbar.
- In the Quick Access Toolbar menu.
- When you right-click a button in the ribbon.

Quick Access Toolbar
47
Find
With an active plot, list or spreadsheet, Find enables finding items meeting selected
criteria.
The Find dialog is opened by clicking Find in the Highlight group on the Tools tab,
by pressing CTRL+F or from the Quick Access Toolbar.


In the table, the available items in the Find in, Find what, and Logical Expression
boxes are listed.
Find in Searches in Logical expressions
Identifiers The text or values of the selected
identifier, primary or secondary, or
all identifiers.
For batches the secondary identifiers
include the batch level IDs Phase,
Maturity, Source, and Number, and
the batch project IDs $BatchID and
$PhaseID when available.
For numerical identifiers:
>, >=, <, <=, between, =, not
equal.
For text identifiers:
begins with, ends with,
contains, =, not equal.
All Data The vectors displayed on all axes.
For the default score scatter plot, the
search is done on both t1 and t2.
>, >=, <, <=, between, =, not
equal.
Series
Data
The vectors that are selected as
series. For 2D scatter, line, and
column plots that means that the
search is limited to the y-axis vectors
while for the 3D scatter plot the
search is limited to the z-axis
vectors. For example, for the default
score scatter plot, the search is
limited to searching in the vector t2
found on the y-axis. For lists, this
option is the same as All Data.
>, >=, <, <=, between, =, not
equal.
X Axis The vector displayed on the x-axis.
Available for plots only.
>, >=, <, <=, between, =, not
equal.
User Guide to SIMCA
48
Find in Searches in Logical expressions
Vectors The selected vector in Find what.
When selecting Vectors in Find in:
- The Find what box contains
all vectors available for the
type of vector displayed in the
active plot or list, e.g. if the plot
displays t1 vs. t2, the available
vectors are all vectors of data
type Variables and Scores.
- The Vector Details section
becomes available enabling
selection of model.
>, >=, <, <=, between, =, not
equal.
Text The list as if all entries where text
and marks the matching cells.
Available for lists and spreadsheets
only.
begins with, ends with,
contains, =, not equal.
Note: To search in a variable, in the Find in box select Vectors and in the
Find what box select XVar. Then in the X Terms box in the Vector details
section, select the desired variable.
Marking type
There are three marking types:
- Clear previous clears the previous marking and marks according to the
current criterion.
- Union with previous marks the items that satisfy the current OR the
previous criterion but not both.
- Intersection with previous marks the items that satisfy both criteria, both
the previous AND the current.
The default marking type is Clear previous.
Open the dockable window Marked Items (select the Marked Items check box on the
View tab) to see the items previously and currently marked. For more, see the Marked
Items subsection in the View chapter.
Quick Access Toolbar
49
Using Find
Here follows a step by step example of using the Find function.
1. Open a plot or list.
2. Click Find in the QAT.
3. In the Find in box, select the category of the attribute.
4. In the Find what box, select the attribute and then specify the criterion.
5. Click the Mark All button and all objects in the plot or list satisfying the
criterion are marked.
6. Then switch criterion and select Marking Type to the union or intersection
with the previously marked objects. The intersection will select among the
objects already marked those that satisfy both criteria. The union marks
objects that satisfy either criterion but not both.

51
File
Introduction
This chapter describes all features available on the File tab.
The following commands are available: New Regular/Batch Project, Open, Save,
Save As, Print, Manage (Encrypt, Compact Project File, Delete Project) Import
Dataset, Generate Report, and Close. The buttons Project Options, SIMCA
Options and Exit SIMCA are available at the bottom of the menu. The right-side of
the menu lists Recent files.

New - Creating a new project
When creating a new project you have to select which type: New Regular Project or
New Batch Project. The first time you create a new project, the default is New
Regular Project. After that the type selected last time a project was created is the
default.

Click File | New Regular/Batch Project to open SIMCA import, select your data and
create a new project. See the SIMCA import chapter for more.
User Guide to SIMCA
52
Open - Opening an existing project
Click File | Open to open an existing SIMCA project by selecting the corresponding
SIMCA project *.usp-file.
Each instance of SIMCA can only open one project while several instances of SIMCA
can be open in parallel.
Opening pre-13 projects
Projects from SIMCA-P version 8.0 and earlier cannot be opened by this version of
SIMCA, but you can import the *.sim file.
Projects from SIMCA-P 9 to SIMCA-P+ 12 are automatically converted when opened.
All models are refitted.
Note: Datasets from SIMCA-P 9 to SIMCA-P+ 12 usp-files can be imported.
Opening pre-13 filtered projects limitations
When converting a filtered project the following functionality is unavailable in
SIMCA:
- Reconstruct.
- Spectral Filters and Time Series Filters.
- File | Import Dataset.
Opening pre-13 batch projects
When opening a batch observation level project, all connected batch level projects are
listed and you can choose which one to convert resulting in a project with both the
observation level models and batch level models. For more, see the Batch chapter.
Batch projects have been reorganized to include both levels in the same project. The
observation level model and dataset have been renamed batch evolution model and
dataset, abbreviated BEM and BE DS, while the batch level model and dataset names
remain batch level model and dataset, abbreviated BLM and BL DS. Both the BEM
and the BLM are positioned in a BMx, where x is a sequential number.
Save project
To save the open project to the current name and location, click File | Save.
Save As
Under Save As you can select to save the active project, plot or list to a specific file
name and path.

File
53
Save Project As
To save the open project to a different name or location than specified when exiting the
SIMCA import click File | Save As | Save Project As.
Save Plot or List As
To save the current plot or list to file, click File | Save As | Save Plot/List As.
Plots can be saved as Enhanced Metafile (*.emf), Bitmap files (*.bmp), Jpeg files
(*.jpg), or PNG files (*.png). 3D plots can additionally be saved as Encapsulated
Postscript EPS (*.eps) and Scalable Vector Graphics SVG (*.svg) but not as (*.emf).
Lists can be saved as Text File (*.txt), DIF Files (*.dif), or Matlab Files (*.mat).
To save a plot:
1. Display it and make sure it is active.
2. Click File | Save As | Save Plot As. The Save Plot dialog is displayed.
3. In the Size box you can choose between
- Custom
- A) 600x375 (Suitable for portrait (A4, US Legal, etc.))
- B) 300x300 (Square, fits two side by side)
- C) 600x600 (Square)
- D) 755x465 (Suitable for PowerPoint, one per slide)
- E) 755x270 (PowerPoint, fits two above and below)
- F) 370x465 (PowerPoint, fits two side by side)
- G) 800x600 (4:3 format)
- H) 1024x768 (4:3 format)
- I) 1280x720 (16:9 format, HD 720)
- J) 1920x1080 (16:9 format, HD 1080)
- Original size
4. In the dialog select the Keep aspect ratio check box to keep the proportions
of the plot as on the screen (Original size) or according to a predefined plot
size.
5. Click OK when done.
6. Select location, file name, and file type. Click Save.

To save a list:
1. Display it and make sure it is active.
2. Click File | Save As | Save List As. The Save dialog is then displayed.
3. Select location, file name, and file type. Click Save.
User Guide to SIMCA
54
Print, Print Preview, and Print Setup
The commands Print, Print Preview, and Print Setup are the standard printing
functions available by clicking File | Print.
Note: Plots are printed as viewed on the screen except when printing to a pdf
writer.
Manage
Encrypt
To encrypt and lock a project, click File | Manage | Encrypt. The Encrypt Project
dialog opens and you are prompted by SIMCA to enter a password.

After entering the password twice and clicking OK, the project is locked and can only
be opened by providing the password.
Compact Project File
When compacting a project file, this is done by removing unnecessary data and
optionally data considered unnecessary by the user. Compacting a project file is
particularly useful when models or datasets have been deleted in the project.
Click the File | Manage | Compact Project File to compact the project file.

In the Compact Project File dialog, select which data to remove. The available
choices are listed in the table.
Note: Compact cannot be undone. The compacted project file is saved when
clicking OK.
File
55
Delete Project
To delete the current project, click File | Manage | Delete Project.

Import
To import more data to the current project, click File | Import Dataset. See the
SIMCA import chapter for how to continue.
For batch projects batch condition data can be imported at the same time as the batch
evolution data or later using Import Dataset.
Merging with another dataset while importing is not necessary as all relevant datasets
can be selected in the Workset dialog and merged by primary ID on the fly when
creating the workset.
Generate Report
Introduction
SIMCA has an automatic report generator available by clicking File | Generate
Report.
In the report generator, basic formatting functionality for writing text is available.
Plots, lists, and model results of SIMCA can be added to the report at any time. These
items are added to the report as placeholders.
A placeholder stands in the place of contents which SIMCA will provide; let it be a
plot, list, text or number.
The placeholders enable using the same report, as a template, in different projects. Edit
the text, and just click Update report to update all SIMCA results using the active
model.
Note: To avoid updating a plot, you must remove the placeholder. For more,
see the Tools menu in report subsection.
For details about how to send a report by e-mail, see the knowledgebase at the
Umetrics web site www.umetrics.com.
User Guide to SIMCA
56
Starting to generate a report
To create a report:
1. Click File | Generate Report.
2. In the Generate Report dialog select to:
a. Create a new report selecting Create new using and selecting
[Blank] or [Umetrics Default Template] or
b. Open a report by selecting Open existing and selecting an existing
report.
3. Click OK.

Generate Report window
The report for the active model opens in a new window, the Generate Report
window. This window is separate from SIMCA but is closed when SIMCA is closed.

File
57
The Generate Report window contains the following:
- Menu bar holding the menus File, Edit, View, Insert, Format, Tools, and
Help.
- Generate report toolbar with commonly used commands.
- Formatting toolbar with commonly used formatting commands.
- Main window showing the report template.
- Placeholders window with a list of built-in placeholders.
- Report Generator FAQ window with a short introduction to how to use the
report generator.
- Properties window where plot size and placeholder properties can be edited.
Default templates
SIMCA has a default template for the following models:
- PCA
- PLS, OPLS, O2PLS
- PLS-DA, OPLS-DA, O2PLS-DA
- Batch
Default template content
The SIMCA default template consists of:
1. Introduction and Objective (text to be filled by user).
2. Description of the included datasets.
3. Pre-processing such as spectral filters etc., if any.
4. Model summary.
5. Transformation and scaling.
6. Goodness of fit.
7. Model results.
8. Predictions (plots and lists to be added by the user).
9. Conclusion (text to be filled by user).
User Guide to SIMCA
58
Appending to, inserting in, or replacing existing report
With an open report, right-clicking a model in the Project Window and clicking
Generate Report opens the following dialog:

This enables:
1. Appending the report updated for the active model to the current report by
clicking append to existing report.
2. Inserting the report updated for the active model to the current report at the
caret position in the report by clicking insert at the caret position in the
report.
3. Replacing the existing report with an updated report using the active model
by clicking replace existing report.
Note: After selecting 1 or 2 above and clicking OK, the placeholders for the
previous report are removed.
Creating a report including several phases
To generate a report for all the phases of the batch evolution model:
- Right-click the BEM in the Project Window.
- Click Generate Report.
- Select the phases to include in the report.
- Click OK to open the report.

File
59
Menu bar in report
The menu bar consists of the File, Edit, View, Insert, Format, Tools, and Help
menus.

File menu in report
Under the File menu the general Windows commands New, Open, Save As, Print
Preview, Print and Exit are available.
Additionally the report generator commands Continue edit with, Templates and View
in browser are available.

General windows commands
The functions of the general windows commands are described in the table.
Command Use to
New Create a new report from the selected template or report.
Open Open a report saved in HTML format.
Save As Save the report in HTML format. The report is saved with
placeholders and can be used as template.
Print, Print Preview Print the report. Preview the report.
Exit Close the report.
Continue editing report with
Continue to work with the report in the editor of your choice by under the File menu
clicking Continue edit with, and then selecting the editor. The applications listed here
are the applications that have registered with Windows that they can edit HTML text.
Continue edit with is also available as a button in the Generate Report bar.

User Guide to SIMCA
60
Templates
Save templates, restore templates and add or remove custom templates by clicking
Templates under the File menu.
- Select Save as default template when you have changed/created a template
according to your wishes and want to use it as the default template next time
you generate a report for the same model type.
- Select Restore default templates if you have made changes to the default
templates and want to remove those changes.
- Select Save as custom template when you have changed/created a template
according to your wishes and want to save it to be able to use it again.
Custom templates will be listed in the Select template or open existing
dialog (File | Generate Report) and when clicking Insert | Template.
- Select Add / Remove custom templates when you want to add an already
created template or remove one of your custom templates.

View in browser
To view the current report in your default internet browser, on the File menu click
View in browser. This works with Internet Explorer and Mozilla Firefox.

Edit menu in report
Under the Edit menu the commands Undo (CTRL+Z), Redo, Copy (CTRL+C), Paste
(CTRL+V), Paste unformatted (CTRL+SHIFT+V), Clear (DELETE), Select All
(CTRL+A), and Find (CTRL+F) are available.
Paste unformatted is useful when the text copied is formatted.
View menu in report
Under the View menu, select to hide or show the Toolbars, Status bar, Placeholders
window, and Properties window. You can also customize the toolbars, command
menus and toolbar and menu options in Customize.
- To show or hide the toolbars click Toolbars and then click Generate
Report or Format. Both toolbars are displayed by default. For more about
the toolbars, see the Generate Report bar and Format bar subsections later
in this chapter.
- To show or hide the Status Bar, click it. The Status Bar displays an
explanation when positioning the cursor over a button.
- Click Placeholders to display the window. For more see the Placeholders
window subsection later in this chapter.
- Click Properties to display the window. The properties of plots and images
can be customized in the Properties window. For more see the Properties
window subsection later in this chapter.
File
61
Insert menu in report
Use the Insert menu to insert a Hyperlink, Image, File, or Template in the current
report. Here you also can toggle the Grab plot or list mode.

Hyperlink
To insert a hyperlink, mark the text or plot in the report and then click Insert |
Hyperlink. The following dialog opens:

In the URL field, enter the hyperlink address.
Insert Hyperlink is also available as a button in the Format bar.

Image
To insert a picture in the report, click Insert | Image, and then click the Browse button
to find the file.

Insert Image is also available as a button in the Format bar.

File
To insert a Web page file (*.htm, *.html), a Text file (*.txt), or a picture file (*.jpg,
*.png, *.gif, *.bmp.), click Insert | File. In the Open dialog, select the file.
User Guide to SIMCA
62
Template
To select a template to insert in the report, click Insert | Template. In the Select
Template dialog, select the custom template to insert in the report.

For a template to be available in this dialog, it must first have been saved using File |
Templates | Save as custom template.
Grab plot or list
To add a plot or list with one click, click Insert | Grab plot or list. This menu toggles
the Grab plot or list mode.
When the Grab plot or list is toggled on, hovering over a plot or list in SIMCA
displays the cursor as a hand . This hand signifies that clicking the plot or list
inserts (grabs) it at the insertion point in the open report.
When done click the menu again to deactivate the Grab mode.
Format menu in report
Use the Format menu to customize Font and Styles and formatting when specifying
your style sheet.

Tools menu in report
Under the Tools menu find the commands Update report, Update placeholder, Show
all placeholders, Show placeholder, Remove all placeholders, and Remove
placeholder.
Command Result after selected
Update report Updates all placeholders with the plots and lists of the
active model. Update report is also available as a button
on the Generate Report bar.
Update placeholder Updates the marked placeholder.
Remove all
placeholders
Removes all placeholders. Use this option when you do
not want any items, plots, or lists to be updated. Remove
all placeholders is also available as a button on the
Generate Report bar.
Remove placeholder Removes the marked placeholder. Use this option when
you do not want a certain item, plot, or list to be updated.
Show all placeholders Shows all underlying placeholders.
File
63
Command Result after selected
Show placeholder Shows the underlying placeholder of the marked plot, list,
or item.

Help menu in report
To access the FAQ of the report generator, click Welcome page and FAQ on the Help
menu. The Report Generator FAQ window opens to the right.
Generate report bar
The Generate Report bar includes the general commands New, Open, Save, Cut,
Copy, Paste, and Undo. All these commands work according to Windows standard
except for the New command described here.
The generate report bar additionally includes the commands Update placeholders,
Remove all placeholders, Grab plot or list, and Continue edit with.

New
Click the arrow next to the New button and the following commands are displayed.

- Click New from the menu displays the generate report dialog from which
you can select which report/template to use or to open an existing
report/template.
- Click New blank report to start a new report with no text.
- Click New from default template to create a report from the default
template. Save a report as a default template by clicking File | Templates |
Save as default template. Umetrics default template is used if no other
template has been specified.
Generate report bar additional buttons
For info about the Placeholder buttons Update report and Remove all placeholders,
see the Tools menu section.
For info about the Grab plot or list button, see the Insert menu section.
User Guide to SIMCA
64
For info about the Continue edit with button, see the File menu section.
For info about the View in browser button, see the File menu section.
Format bar
The Format bar is the standard toolbar for formatting text with three additional
buttons: Insert hyperlink , Insert image , and View in browser .
For more, see the Insert menu and File menu sections in this chapter.
Placeholders window
The placeholder window is accessible through the menu View | Placeholders.
The placeholders are organized into the following categories:
- General
- Model
- Templates

Inserting a placeholder
To insert a placeholder in the report:
1. Place the cursor in the desired position in the report.
2. Mark the placeholder by clicking it in the Placeholders window.
3. Click Insert.
Placeholders window - General
The general placeholders available are listed here with the expected result after
clicking Insert and updating the placeholder.
General HTML
- Horizontal line - horizontal line.
- Hard line break - line break as when pressing SHIFT + ENTER.
General
- Application name - 'SIMCA'.
- Application version - 13'.
- User name - the user name used to log in to Windows.
- Computer name - the internal computer name.
- Current date - today's date.
- Current time - the time when updating the placeholder.
- Project title - the name of the SIMCA project.
File
65
- Project file name - the SIMCA project name including extension, e.g.
FOODS.usp.
- Project path - the path to the SIMCA project file.

Placeholders window - Model
The model placeholders available are listed here with the result after clicking Insert
and updating the placeholder. The placeholders refer to all included datasets.
Numbers
- Num. variables - number of variables included in the model when counting
both X and Y.
- Num. X variables - number of x-variables in the model.
- Num. Y variables - number of y-variables in the model.
- Num. excluded variables - number of excluded variables.
- Num. observations - number of observations in the model.
- Num. excluded observations - number of excluded observations.
- Num. batches - number of batches in the model. Batch projects only.
- Num. classes - number of classes.
Text
- Variable primary ID - lists all primary variable identifiers in the model.
- X variable primary ID - lists the primary variable identifiers for x-variables
in the model.
- Y variable primary ID - lists the primary variable identifiers for all y-
variables in the model.
- Excluded variable primary ID - lists the primary variable identifiers for
excluded variables.
- Included batch ID - lists the batch identifiers of the batches included in the
model. Batch projects only.
User Guide to SIMCA
66
- Excluded batch ID - lists the batch identifiers of excluded batches. Batch
projects only.
- Observation primary ID - lists the primary observation identifiers in the
model.
- Model name - lists the model name, e.g. M1. For batches, with the phase
name, e.g. M1:chip.
- Model title - lists the user entered model title. 'Untitled' if no title was
entered.
- Model type - lists the model type, e.g. PCA. For batches with the phase
name, e.g. PLS-Class(chip).
- Phase ID - lists the phase identifier of the active model.
- Dataset names - lists the names of all datasets included in the model.
Summaries
- Dataset Import Log - the text from the import log. Can be viewed in
SIMCA by opening the Dataset Properties, tab General.
- Preprocessing Summary - the text in Data | Dataset Summary. Available
for a filtered datasets.
- Expanded terms - lists all expansions.
- Lagged variables - lists all lags.
- Transformed variables summary - lists the transformed variables and their
transformation.
- Scaled variables summary - lists the variables not centered and scaled to
Unit Variance with their scaling.
- Trimmed variables summary - lists the trimming performed in the workset.
Results
- Num. Components - lists the number of components extracted for the active
model.
- R2X, R2X(Cum), R2Y, R2Y(Cum), Q2, Q2(Cum) - lists the values for all
components or cumulatively for the last component for the vectors. For
details about the vectors, see the Function of component vectors section in
Appendix A: Statistics.
Note: The vectors are by default displayed for the last component. For how to
display all components, see the Properties window
Multi-plots
- Auto-generated XVar plots - displays each x-variable in a separate line
plot.
- Auto-generated XObs plots - displays each observation in a separate line
plot. Only the values for the x-variables are displayed.
Using these placeholders it may take a long time to update the report. Press the ESC-
button to abort the multi plot update.
File
67
Placeholders window - Templates
The Templates placeholders consist of model templates.
Use the templates placeholders to insert a template into a report.

Properties window
The properties window is opened by clicking View | Properties. The properties for a
placeholder are displayed when clicking the placeholder with the Properties window
open.

In the Properties window the following can be customized:
- The Default plot size and format under Default Settings.
- The Plot size and format for the current plot (placeholder) under
Placeholder Properties.
The format for plots is by default .jpg. The available formats are .png, .bmp, and .jpg.
Placeholders for vectors can be modified to display a part of the vector or the entire
vector. For instance, R2X is by default displayed for the last component. To display all
components, type: 'R2X[{0}] = {1}{separator: }<BR>' in the Data field.
Note: The properties of the placeholder for the date can be changed by
selecting it and using the Properties window.
User Guide to SIMCA
68
Adding plots and lists to the report
To add a plot or list to the report, click the desired position in the report and then use
one of the following methods:
- Right-click the plot or list window in SIMCA and click Add to Report.
- On the View tab, in the Add group, click Add to Report.
- Click the Grab Plot or List-button in the Generate Report window and
click the plot or list.
The plot or list is inserted into your report.
Close project
To close the current project and all its windows, click File | Close.
Recent files list
Recent files lists the most recently opened files.
Clicking the pin to the right of the recently opened file makes the pin green and keeps
that file in the recent file list even after it should have fallen off.

File
69
Project Options
Options in SIMCA can be set at 3 levels:
- Model Options set at the model level and are local to that particular
model. Some options can only be set in model options. New models inherit
the options from the project level. For more about the model options, see the
Model Options section in the Workset section of the Home chapter.
- Project Options set at the project level and are in effect for that project
and all models created after applying changes. New projects inherit the
factory settings options. To make new projects inherit the current project
options settings, click the Save as General button in the General tab.
- SIMCA Options - apply to the software. Changes in these options apply to
all projects opened after the change. For more about the general options, see
the SIMCA Options section later in this chapter.
This section describes the Project Options dialog available as a button below the
recent file list on the File tab.
The Project Options dialog includes the tabs: Fit, Predictions, Distance to Model,
Plot labels, Residuals/R2, Performance, Audit Trail and General.
Fitting options
The options available from the Fit page pertain to the fit of the model.


All options are described in the table below:
Option Description Default
Cross
validation
rounds
Each observation is left out once during the
cross validation. The number in Cross
validation rounds is the number of groups, that
one by one is left out of the modeling and
repredicted during the cross validation.
7
User Guide to SIMCA
70
Option Description Default
Max
iterations
Maximum number of iterations until
convergence when fitting the model.
200
Missing
data
tolerance
(%)
Threshold of missing values for observations
and variables. The threshold applies to both the
workset and the predictionset.
The threshold is displayed in the Workset
dialog and can be changed there for the specific
model.
50%
Min number
of non-
median
values
Minimum number of values that have to differ
from the median. When a variable has fewer
values differing from the median than entered in
Min number of non-median values, SIMCA
will suggest excluding the variable since it is
considered to lack variance.
3
Confidence
level on
parameters
Confidence level used when computing
confidence intervals on the parameters. None,
99%, 95% and 90% are available.
95%
Significance
level
Significance level used to compute the
Hotellings T
2
ellipse and the critical distance to
the model.
0.05 (95%)
Default
scaling
Default scaling for X-block respective Y-block in
the default workset. Changes here apply to the
x-variables and y-variables defined as such in
the default workset.
UV
Reconstruct
wavelets
When using the wavelet transform to compress
the dataset variable wise (suitable for spectra
such as NIR, or Raman etc.), SIMCA creates a
new dataset and you can fit models to the
compressed data.
When variables have been compressed, the
new variables are linear combinations of the
original ones. Loading, coefficients, VIP or any
plots displaying variables are difficult to interpret
in the wavelet domain. Therefore it is possible to
reconstruct not only the original variables, but
also individual vectors such as
loadings, coefficients, VIP, etc.
By default
the
Reconstruct
wavelets
check box is
selected,
and all plot
are
displayed in
the domain
of the
original
dataset.
Reset Clicking the Reset button returns all settings to
the factory settings for this Project Options
page.

Note: To not display the Hotellings T2 ellipse or the D-Crit line, see the
Limits and Line Style subsections in the Plot and list contextual tabs chapter.
File
71
Predictions presentation
The options available from the Predictions page pertain to the presentation of the
predictions.

Transforming predictions
When the y-variables have been transformed, by default the predictions are back
transformed to the original units. Select the Transform predictions check box when
you want to display the predicted Y in the transformed units.
Scaling predictions
To display the predicted Y in the same unit as the workset, select the Scale predictions
check box.
Trimming predictions as the workset
When the workset has been trimmed or Winsorized, the predictionset can be trimmed
or Winsorized in the same manner by selecting the Trim predictions as the workset
check box.
By default the predictionset is not trimmed nor Winsorized.
Distance to the model plot options
The options available from the Distance to Model page pertain to the presentation of
the distance to model plots.

User Guide to SIMCA
72
Normalized or absolute distances
The distance to the model of X or Y, DModX or DModY, can be expressed as an
absolute value or a normalized value i.e., in units of standard deviation of the pooled
RSD of the model.
The default is Normalized in units of standard deviation.
Weighted by the modeling power
When computing the distance to the model, SIMCA by default does not weight the
residuals by the modeling power of the variables. Change this default by selecting the
Weighted by the modeling power check box. The distance to the model will then be
computed with weighted residuals.
Selecting plot labels
The options available from the Plot Labels page pertain to the presentation of labels in
the plots.

Select the identifier to be used as variable or observation label on all plots in the
current project in the Select default variable/observation labels sections. Change the
starting position and length as needed.
To display a specific label type in the title of plots, select the identifier in the Title
variable/observation labels box.
These options are also available from the Properties dialog for the individual plots.
File
73
Selecting type of residuals and R2
The options available from the Residuals / R2 page pertain to the presentation of
residuals and R2 in plots.

Raw or standardized residuals
The normal probability plot of residuals is available on the Analyze tab by clicking
Residuals N-Plot. This plot can be displayed both in original units and in standardized
units. The standardized residuals are the unscaled residuals divided by their standard
deviation.
Select:
- Raw original units to display the unscaled residuals.
- Standardized to display the standardized residuals (default).
R
2

R2 is displayed in all summary plots but the OPLS/O2PLS Overview plot.
Select:
- R
2
- explained variation to display the fraction of the Sum of Squares (SS)
explained by the model (default).
- R
2
Adjusted - variance to display the fraction of variance explained by the
model, SS adjusted for the degrees of freedom.
User Guide to SIMCA
74
Performance improvement using data caching
From the Performance page two types of caching is available:
- Cache calculations in memory - by default selected and caches calculations
to speed up the computations.
- Cache dendrograms in the project file - by default selected and caches
dendrograms to speed up regeneration of the dendrogram.
Clear these check boxes if the project takes too much RAM. The software will then be
slower but the memory consumption smaller.

Administration of the audit trail
The Audit Trail page holds settings and options pertaining to the audit trail.
The audit trail can be turned on/off in SIMCA but for the audit trail to be
automatically on or off for new projects it has to be turned on/off in the
SIMCA import Options available on the File tab.
An administrator can enable or disable the entire Audit Trail page and control the
behavior of the Audit Trail, i.e., always on or off.
As the administrator of the system, contact your Umetrics sales office for instructions
on how to disable the Audit Trail page or search the knowledgebase at
www.umetrics.com.

File
75
The options available from the Audit Trail page are described in the table below:
Option/Button Description Default
Log events to the
Audit Trail
When selected, the Audit
Trail window logs all events
in a session.
For details about the audit
trail, see the Audit Trail
section in the View chapter.
Note that to log all actions in
the SIMCA import in the
audit trail, you have to select
Yes in the Log Events to
the Audit trail in Options in
the SIMCA import before
opening the first dataset.
By default the audit trail is
disabled.
Clear Audit Trail Clears the audit trail for the
current project when clicked.

Prompt new
users to edit the
extended user
information
When selected, new users
are prompted to enter
additional information.
The Extended User Info is
available in the More
Options tab in File |
SIMCA Options.
By default not selected.
Save Saves the audit trail to .xml
when clicked.

User Guide to SIMCA
76
General
The General tab in Project Options allows you to save the current project options as
global options or reset to factory settings.
Save as General Options
Changes in Project Options apply to the current project only. To use the current
modified project options for new projects created in the future, click the Save as
General button in the General tab, and click Yes to verify.

Reset All Options
To revert all changes in the Project Options dialog to the default (factory) settings,
click the Reset All button in the General tab.
File
77
SIMCA Options
The options available from SIMCA Options apply to SIMCA and not just the current
project.
The tabs available are General, Threading, Restore, Painter, and More Options.
These pages are described in this section.
SIMCA Options is available from the File tab, below the recent file list.
General
The General page includes default behavior when starting SIMCA, creating a new
project, and fitting the model.

All check boxes, starting from the top, are described in the table below:
Option Description When Default
Open last
opened
project on
start-up
Opens the last
opened project
when opening
SIMCA.
The project you worked with
last is often the one you start
off with.
Clear this check box when
working with very large
projects, to avoid opening a
project when planning to work
with another.
By
default
selected.
Close the
already
open
project
when
opening a
new project
The open project
will be closed and
the new created
one will be the
only open.
When you want only one
project open.
Clear this check box if you want
the current project to remain
open. The new project is then
opened in another instance of
SIMCA.
By
default
selected
Show the
summary
plot while
fitting
When fitting a
model, the
summary plot is
displayed.
Clear the check box if you don't
want to see the summary plot.
By
default
selected.
User Guide to SIMCA
78
Selecting multithreading
In SIMCA certain calculations can use multithreading to speed up the calculations, if it
detects that the computer has several processors or several processor cores. Threading
can speed up calculations for large projects but will have little effect on small to
medium sized projects.
In the Threading page of the File | SIMCA Options dialog, threading can be turned
on and off and the number of processors to use can be specified.

Using multithreading
In the Use multithreading section, select one of the following options:
- Default SIMCA checks if the computer has more than one processor and
turns multithreading on if it does. In parenthesis the status is displayed as
currently using or currently not using.
- Always use multithreading SIMCA uses multithreading without first
checking for number of processors.
- Never use multithreading SIMCA does not use multithreading.
Selecting number of threads
In the Maximum number of threads section, select one of the following options:
- Default - SIMCA threads using the number of processors found. That
number is displayed in parenthesis.
- Use x threads; enter the number of threads to use - This option is only viable
when you want fewer threads than available processors to be used.
File
79
Restoring to default
The Restore page in the SIMCA Options dialog can be used to restore SIMCA
defaults by clicking the respective Restore buttons.

The table describes the available restoring options.
Restore Description Restoring is useful when
Format
Plot
A file defining how all plots
look, including fonts,
colors, plot marks etc.
See also the Switching
plot formatting templates
topic in the Layout
chapter.
The changes saved in the default
plot template are no longer
desired.
Favorites A dockable window
displaying shortcuts to
plots and lists.
The original Favorites items are
desired.
Windows
Positions
The positions of dialogs
and windows are saved.
A dialog ends up outside the
screen, for instance when
working with two screens and
one screen is no longer available.
Don't show
again
messages
A Don't show again-
message is an information
dialog with a Dont show
again check box that if
selected hides the
message the next time it
could be displayed.
You want to know what a Don't
show again message said.

User Guide to SIMCA
80
Selecting progress bar pictures
The Painter page in the SIMCA Options dialog provides the option to display nothing
or the work of a painter during fit of models. Under Progress Bar select your
Preferred painter.
Painters possible to select are Ando Hiroshige, Vincent van Gogh, William Turner,
Diego Rivera and Pierre-Auguste Renoir. Selecting Random displays the work of all
artists in random order. Selecting None turns this feature off.

To create a collection of your own images to display on the progress bar, download the
ArtBuilder from the support knowledge base at www.umetrics.com and follow the
instructions included.
More Options
The More Options page displays and holds settings for default directories, plot
options, printer options, tweaks, and the audit trail.

In the table the options available in the More Options page are described. Each option
can be changed by clicking the current entry and changing it.
File
81
Option name Description Default
Default Directories
Temporary
Directory
The path to the Temporary Directory, used by
SIMCA to save temporary files, is displayed.
A
temporary
directory in
the
Windows
default
temporary
directory.
Plugin Directory The path to the Plugin Directory, used by
SIMCA to load plugins, is displayed. Plugins
supported by SIMCA are generated variables,
spectral filters, import, and item information.
A plugin
folder in
the
application
data folder
of the
current
user.
Plot Options
Remember
coloring
Remembers the coloring specified in Tools |
Color by or in the Color tab of the Properties
dialog, between plots. That is, after selecting a
certain coloring in one plot, the next plot
opened will use the same coloring scheme.
Yes
Remember
sizing
Remembers the sizing specified in Tools | Size
by or in the Size tab of the Properties dialog,
between plots. That is, after selecting a certain
sizing in one plot, the next plot opened will use
the same sizing.
No
Remember item
selection
Remembers the item selection specified by
marking and clicking Hide on the Marked
Items tab, or the Item Selection tab of the
Properties dialog, between plots. That is, after
making the item selection for one plot, the next
plot opened will use the same item selection.
No
Remember label
ID
Remembers the label specified in Tools |
Labels or on the Plot Labels tab in the
Properties dialog.
Yes
Plot labels limit When the number of plot labels in a plot
exceeds this number, no labels are displayed.
500
Automatically
change active
model
The active model is automatically changed
when clicking a plot created from that plot.
No
Plot engine Switch between GDI+ and Direct 2D. Direct 2D
is not supported by Windows XP and Windows
Vista might need an update.
Direct2D
when
possible.
Printer Options
User Guide to SIMCA
82
Option name Description Default
Print lists in
color
Colored lists will be printed in color on a color
printer, or in different shades of grey on a
regular printer. When Print lists in color is No
the lists are printed black and white.
Yes
Tweaks
Automatic find
delay time
(millisec.)
When typing in an edit-field, for instance the
Find field in the Workset dialog or the X
Terms field when selecting XVar in the
Plot/List | Scatter Plot dialog, SIMCA
searches the list for a match. The time before
starting this automatic find is defined in
Automatic find delay time (millisec.).
500
Audit Trail
Prompt for
Extended User
Info
When set to Yes new users opening a project
with audit trail turned on will be prompted to
enter more user information than the default.
The entered information can be viewed in
Extended User Info in this dialog.
No
Extended User
Info
User entered information to be displayed
alongside the default user information in the
audit trail.
<empty>
Correlation Matrix
Correlation
matrix limit
Increasing from the default allows displaying
the correlations of more variables.
32
Exit
To close SIMCA, click File | Exit.

83
SIMCA import
Introduction
When creating a new project or selecting to import another dataset, the SIMCA
import opens allowing you to import from file, database or just paste the data in an
empty spreadsheet.
The first time you open the SIMCA import, the Open dialog for selecting a file is
opened.
The SIMCA import window consists of the ribbon, the quick access toolbar at the top,
dockable windows and the Find toolbar.

User Guide to SIMCA
84
File
This section describes all features available from the File tab in SIMCA import.
The following commands are available: New Spreadsheet, Open Workspace, Save
Workspace, Save As, Print, and Close. The buttons Options and Cancel are
available at the bottom of the tab. The right-side of the tab lists Recent files.

New Spreadsheet
The SIMCA import supports three types of import under New Spreadsheet:
- From File - Opens a dialog where you can select file to import.
- From Database. For more see the Importing data from a database
subsection later in this chapter.
- Blank - enables pasting in an empty spreadsheet. For details, see the Pasting
dataset in import wizard subsection later in this chapter.
Select any type of import by clicking File | New Spreadsheet | From File/From
Database/Blank. Repeat until all data you want to import are available in SIMCA
import.

Note: The import type used is the default the next time you import.
SIMCA import
85
Import from file
When selecting to import from file, the Open dialog opens. This dialog is a standard
dialog box for selecting file type, name and source address of the data file to import.
The selected file type in the Files of type box is the default file type next time a file is
imported.
Note: The datasets in .usp files can be imported.
In this dialog you can also select to downsize the dataset before reading by selecting
the Downsize imported data check box, positioned bottom left. For details, see the
Downsizing dataset before reading it subsection later in this chapter.
File types supported
In the Files of type list box, select your file type. SIMCA imports a number of file
types. See the table.
Supported Files is the default selected file type and results in that all files of the
supported file types are listed.

Note: For JCAMP-DX, NetCDF, Galactic SPC, and Brookside XML files,
importing several files from the same folder, and merging without (exact)
matching, is available by marking them and clicking Open.
Downsizing dataset before reading it
Reading large datasets may be time consuming. In the case where you plan to
downsize the dataset, selecting to downsize before reading the dataset can reduce the
time.
To downsize the dataset before reading it:
1. In the Open dialog, select the Downsize imported data check box, select
the file, and then click Open.
2. In the Downsize dialog:
- In the Start after xx rows field, enter the last row to read before
downsizing.
- Enter the downsizing interval in the Keep every xx rows.
3. Click OK when done.
User Guide to SIMCA
86

Downsizing is also available after opening the file in SIMCA import. For more, see the
Downsizing the dataset subsection later in this chapter.
Importing data from a database
To be able to use the database import utility in SIMCA, a data source and an interface
to the data source already needs to be in place.
There are three main interface types supported:
- ODBC (Microsoft)
- OPC (OPC Foundation)
- SimApis.
The SimApi is the API (Application Program Interface) connection that SIMCA-online
needs to be able to connect to data sources. For instance OSIsoft PI and Siemens
SIPAT have support for the SimApi interfaces.
The database import supports several extraction methods:
- Continuous - from a start time to an end time.
- Batch - batches that exists in the system during a time interval.
- Relational - a complete table in an ODBC data source.
- SQL - an SQL query through the ODBC driver from the database vendor.
In table 1, the different possible import combinations are shown.
Table1: Different combinations of data connections and retrieval types supported.
Continuous Batch Relational SQL
ODBC X X X X
OPC X
SimApi X X
Since the configuration of some of these different combinations can be rather
cumbersome, a SIMCA DataBase Settings file (*.sdbs) can store most parts of the
configuration.
For more information about the database settings file, see the SIMCA Database
Settings file subsection later in this section.
SIMCA import
87
Manage Data Sources
In the Setup Data Sources page of the Database Import Wizard all data source
connections are set up.

Use the buttons above the data source list to:
- Add new ( ),
- Remove ( ) and
- Test ( ) a data source connection.
Data sources connection information is stored in the registry. There are three types of
connections possible:
- ODBC ( )
- OPC ( )
- SimApi ( )
Unavailable connections appear gray.
An example: The picture below displays an available SimApi data source, with three
nodes, that is selected as one of the data sources to use when importing the data.

When a settings file is available, it can be selected in the recent list, browsed to, or
entered in the Database Settings file box.

User Guide to SIMCA
88
ODBC - Add New Data Source
To create an ODBC connection:
1. Click Add new in the Setup Data Sources page.
2. In the Add new data source dialog, select ODBC in Connection type box.
3. Select the data source name (DSN) that you have set up in the Windows
ODBC Data Source Administrator application.
Note: Configuration of ODBC sources are driver specific, but can for
instance contain log in information and server information.

OPC - Add New Data Source
To create an OPC connection:
1. Click Add new in the Setup Data Sources page.
2. In the Add new data source dialog, select OPC in the Connection type box.
3. Browse to the network computer to connect to.
4. Expand the network node to view all available servers on that computer.
OPC servers with the historical data interface (HDA) are selectable.

SIMCA import
89
SimApi - Add New Data Source
To create a SimApi connection:
1. Click Add new in the Setup Data Sources page.
2. In the Add new data source dialog, select SimApi in the Connection type
box.
3. Browse for the SimApi DLL file.
Note: The configuration is specific to each SimApi.

Configure Data Source
In the Configure Data Source page, the type of import (Continuous, Batch or
Relational) and necessary tag (variable) definitions are specified. The Current
configurations window to the right (see figure) will help and guide you through the
configuration of the data source.
The toolbar below consists of the following commands: collapse/expand node,
continuous/batch/relational, sampling interval/set time stamp tag, set batch
node/batch ID tag/start tag/stop tag.

Figure A displays a SimApi that is set up as a Batch import with:
- BAKERS YEAST BATCH as batch node ( )
- BAKERS YEAST as the process node ( )
For the process node, the Batch ID ( ) has to be specified.
The exclamation mark on the Data Source icon indicates that the configuration is
incomplete. In the red text to the right we can see that the Sampling Interval to import
data with has not been specified, and that the node EXPORT2 is not configured ( ).
Before continuing the sampling interval has to be specified by clicking the Sampling
Interval button.
User Guide to SIMCA
90
Figure A. The Configure Data Source dialog for a batch import.

Clicking the Sampling Interval button opens the Sampling Interval dialog. Enter the
sampling interval to read data for the marked data source.
Figure B: An OPC data source with two nodes set up correctly and an ODBC data
source configured as a relational import.

Select Tags to Import
The tags, or variables, that will be imported are selected on the Select Tags to Import
page. Also, the user has the possibility to set aliases on the tag names, group the tags in
folders ( ) or add a data filter ( ) with a condition/logical expression on the tags
that will be imported.
In this page, SQL queries can be entered by adding a SQL query folder ( ). SQL
(Structured Query Language) is a programming language to manage data in a database,
however in Simpler we open the database as read only, which means that no alterations
of the database will be introduced in the original database.
SIMCA import
91
In figure C, the alias MyAlias was added to the tag (variable) BATCH ID. The same
tag is also the batch identifier which can be recognized by its icon. The node
BAKERS YEAST contains 8 tags and is a process node and the node BAKERS
YEAST BATCH also contains 8 tags and is a batch node.
Figure C. The Select Tags to Import page with alias specified.

Bottom left in the dialog there is a text filter field enabling text search in the Available
tags section of the dialog. The text filter searches in node names, in tag names and in
aliases.
In figure D, the word "BATCH" was used to filter the available tags and in the
Selected tags for import part, two tags have been added. Note that these tags have the
same name, but are different as they belong to different nodes. The tooltip for the tags
contains node information.
Figure D. The Select Tags to Import page after adding tags to import and with the text
filter for "Batch".

User Guide to SIMCA
92
Data Filters
When adding a data filter, you have to specify a filter name, the logical expression for
filtering and optionally a sampling interval. The data imported according to the filter
are added to the SIMCA import in a new sheet named as the filter.
If the Remove filtered tags check box is cleared, all data will be imported from the
database and a new variable specifying which data, among these data, that satisfy the
filter.
When specifying the filter, each tag is referred to using the number in the dialog, not
the name. When double-clicked, the tag number ends up in the Filter expression field.
In the example in figure e, the filter selects data where the level of MOLASSES is
below 4.75. Note also that time interval for this variable is different from the default.
Figure e: The Add or edit a data filter dialog.

After the filter has been added, tags have to be added under that filter. The added tags
specify which data to extract using the filter.
Figure f: Select Tags to Import dialog with tags added to the filter.

SIMCA import
93
Select Observation Interval
When the data import type is Continuous or Batch, the time interval for which to
import observations and batches can be selected in the Select Observation Interval
page.
The selected Start and End times specify the time interval:
- The observations are imported for in the Continuous case.
- The batches are displayed for in the Batch case. Among the displayed
batches, you can select to include all (default) or some by clearing the
relevant check boxes.
Clicking the padlock locks the time interval so that changing either of the Start or End
times automatically changes the other keeping the time interval constant. Clicking the
arrow moves the date and time in End to Start. This means that by clicking the
padlock you can specify a new time interval, spanning the same time and starting
where the last interval ended, by just clicking the arrow.
In the example below there are several batches between the start and the end time and
all are currently selected for import.
To select only a few batches, clear the Select all batches check box and then select the
batches to import.

User Guide to SIMCA
94
Summary page
The last page in the database import wizard, the Summary page, provides information
about the import. In the Summary page you can save all configurations made into a
new or existing settings file by clicking the Save button to the right of the Database
Settings file field.
The example below shows the Summary page for a batch import. No settings file was
selected.
To save the settings, click Save and enter the file name for the settings file.

SIMCA Data Base Settings file
The *.sdbs file will store the following settings:
- Data Source Configurations
- Aliases
- Filter information
- Tags
- Folders
- SQL queries
Settings not stored in the settings file, but in the registry of the local machine are:
- Data sources.
- Start and stop time in the Select observation interval page (if applicable).
- Recent SIMCA database settings files.
Note: Batches found and/or selected in the Select observation interval page
are not stored anywhere.
SIMCA import
95
Compatibility tested
The database import has been tested at Umetrics using the data source connectors listed
in this subsection.
ODBC Drivers Compatibility
The following ODBC drivers have been tested:
- SQL Server 6.01
- PostgreSQL ANSI 9.0
- MySQL ODBC Driver 5.1
- Microsoft Excel/Access Driver 12.0
- Aegis ODBC Driver
OPC Compatibility
The SIMCA import can only read data through the HDA 1.1 interface of the OPC
Foundation specifications. To see the compatibility of each server, please visit the OPC
Foundation web page (www.opcfoundation.org).
SimApi Compatibility
The following SimApis have been tested:
- OPC SimApi2
- OPC/ODBC SimApi2
- PISimApi2
- ODBC SimApi2
Pasting dataset in SIMCA import
To paste data in the SIMCA import:
1. Click File | New Spreadsheet | Blank. If a dialog is opened when the
SIMCA import is opened, click Cancel to exit the dialog but not the import.
2. In the spreadsheet that opens, paste the data to import.
Note: In the empty spreadsheet you can type new values using the keyboard.
Importing another dataset
To import another dataset in SIMCA import, click File | New Spreadsheet or on the
Home tab, in the Import group, click Import Another.
For details about how to continue, see the New Spreadsheet subsection earlier in this
chapter.
To merge with a dataset already open in SIMCA import, see the Merging
Spreadsheets subsection later in this chapter.
User Guide to SIMCA
96
Workspace
In SIMCA import you can open and save workspaces. A workspace consists of one or
more data spreadsheets formatted as currently specified.
A workspace is saved as a .wusp and can be opened by the SIMCA import.
To save a workspace, click File | Save Workspace.
To open a workspace, click File | Open Workspace.
Save As in SIMCA import
The current spreadsheet can be saved by clicking File | Save As | Save Spreadsheet
As. The file types available are .txt and .dif.
It is also possible to create the SIMCA project from this menu item by clicking File |
Save As | Finish Import. The SIMCA import is then closed and SIMCA opens.
Print, Print Preview, and Print Setup
The commands Print, Print Preview, and Print Setup are the standard printing
functions available by clicking File | Print.
Note: Plots are printed as viewed on the screen except when printing to a pdf
writer.
Close
To close the current spreadsheet in SIMCA import, click the tab cross of that
spreadsheet tab or click File | Close.
SIMCA import
97
Options
Clicking File | Options button opens the SIMCA import Options dialog.


Header What and why Default setting
Undo
You can select to turn the Undo memory
off to speed up spreadsheet actions
involving a lot of data.
On/Off is On.
Undo Actions is 10.
Audit Trail
Turning the audit trail on in the SIMCA
import makes the audit trail include
changes done during import. Optionally
more user info can be entered here. See
the Audit Trail section for more.
Log events to the Audit
Trail is No.
Extended User
Information is empty.
Missing
Value
See the Missing value representation
subsection for details.
Missing Value is empty.
Create
Default
Workset
When creating the SIMCA project a
default workset can be created. When
the default workset is not of interest, the
creation of the default workset can be
turned off.
Create Default Workset
is Yes.
Default
Directories
By default the directory where you select
a data file to import is used for storing
the project.
If you store your SIMCA project files in a
location different from where you fetch
data for import, select Yes.
Enable different default
directories is No

User Guide to SIMCA
98
Missing value representation
Missing values are represented by blank or space < >.
You can select another value to also represent missing in Options dialog in the
SIMCA import by entering it in the Missing Value field.
Note: 99 is assumed to be a valid value and is therefore automatically
changed to 99.0001. If -99 represents missing value, enter -99.0001 in the
Missing Value field in Options dialog in the SIMCA import.
For more, see the Conventions section in the Overview of SIMCA chapter.
Quick Access Toolbar
Standard editing commands Undo, Redo, Cut, Copy, and Paste are available from the
Quick Access Toolbar in the SIMCA import. Any button in the SIMCA import can be
added to this toolbar.

For details about how to customize the Quick Access Toolbar etc, see the Quick
Access Toolbar chapter earlier.
Find toolbar
The Find toolbar opens when pressing CTRL+F or when selecting it from the Quick
Access Toolbar menu.


Using the Find toolbar you can find cells matching a certain expression, numerical or
text.
To find:
1. In the first drop down menu select
- Containing (default),
- Exactly Matching,
- Not Containing, or
- Not Matching that all apply to both text and values
SIMCA import
99
- Less than,
- Greater than,
- Less than or Equal to,
- Greater than or Equal to, or
- Between which all apply to values.
2. Enter the value or text in the first field.
3. In the second drop down menu select:
- Selection - to only search in the current selection.
- Active Document - to search the current spreadsheet.
- All Documents - to search all spreadsheets open in SIMCA
import.
4. If you want to replace, select the Replace check box and enter the new
text/value in the Replace with field. When selecting the Replace check box
two new buttons appear Replace and Replace All. Use these to replace once
the items to replace have been found.
5. Click the arrows to find from top left to right going down (down arrow) or
from bottom right to left going up.
6. Click the Mark All button to mark all cells that match.
7. In Options select:
- Wildcards to be able to use the wild characters ? to represent an
unknown character or * to represent unknown beginning or
ending of the expression.
- Regular Expressions to interpret the search string as a POSIX
basic regular expression.
- Match Case to match character case.
Note: Wildcards or Regular Expressions can be selected, not both. Match
Case can be selected in combination with Wildcards and Regular
Expressions.
User Guide to SIMCA
100
Home tab
In the data spreadsheets open in SIMCA import, all identifiers, qualitative variables,
time or maturity (y-variable for batch projects), and batch condition variables must be
identified and specified. Optionally, regular projects can specify y-variables and
classes defining the default workset.
All column formatting can be performed either from using the ribbon or by clicking the
column or row arrow.
Note: A row, with slash </> as the leftmost character of the leftmost column
of that row, is automatically marked as excluded by SIMCA.
A column with slash </> as the leftmost character in the top cell, is
automatically marked as excluded by SIMCA.

Specifying identifiers
Primary variable and observation identifiers (IDs) are used by SIMCA for the
following tasks:
1. To keep track of variables and observations.
2. Variable and observation IDs are displayed in plots and lists.
3. The Find function in the Workset dialog can search in the identifiers.
4. In the Observations page of the Workset dialog the identifiers can be used
to set classes.
In the import spreadsheet the primary IDs, and batch and phase IDs for batch projects,
must be specified. These IDs can be specified in the ribbon and in the row/column
arrows; see the How to specify identifiers subsection next.
Class IDs can also be specified resulting in classes in the default workset. For more,
see the Class ID specification subsection later in this chapter.
Note 1: IDs in plots are by default displayed using the first 10 characters.
Note 2: SIMCA is case insensitive.
SIMCA import
101
How to specify identifiers
In the import spreadsheet the primary and secondary identifiers, and the batch and
phase identifiers, can all be specified in as described in the table, after marking the
relevant rows or columns. Batch and Phase IDs can only be specified in batch datasets.
Description Illustration
1. Click the
relevant ID button
in the Identifiers
group.

Marked column


Marked row

2. Select ID from
the column
respective row
menu.

Note: If no row is specified as primary variable ID, SIMCA automatically
creates a primary variable ID as Var_1, Var_2, etc., when clicking Finish.

Note: If no column is specified as primary observation ID, SIMCA
automatically creates a primary observation ID as 1, 2, 3, etc., when clicking
Finish.
User Guide to SIMCA
102
Variable and observation ID description
The identifiers, available in SIMCA, are described in the table. Variable IDs are found
in rows and observation IDs in columns.
ID Description Default
Primary
Variable ID
(Identifier)
The Primary Variable ID is required
and needs to be unique.
The primary variable ID row is colored
dark green.
The first row with
unique entries is
automatically specified
as the Primary
Variable ID.
Secondary
Variable ID
Secondary variable identifiers are
optional and colored in light green.
No secondary variable
IDs are automatically
specified.
Primary
Observation
ID
The Primary Observation ID is
required and needs to be unique.
The primary observation ID column is
colored dark yellow.
The first column with
unique entries is
automatically specified
as the Primary
Observation ID.
Secondary
Observation
ID
Secondary observation identifiers are
optional and colored light yellow.
The second column is
automatically specified
as secondary
observation IDs when
the first column is
numerical and the
second text.
Class ID Specifying the Class ID at import
defines classes in the default workset.
Only one column can be defined as
Class ID.
Class ID is only available for regular
projects.
No class identifiers are
automatically specified.
Batch and
Phase ID
When the dataset is a batch evolution
dataset, batch identifiers and
optionally phase identifiers have to be
specified. These specify the start and
end of batches and phases.
Batch IDs are colored orange and
Phase IDs light orange.
Batch and Phase IDs are found in
columns.
There can only be one column defined
as Batch ID and one defined as Phase
ID.
When there are
columns named 'Batch
ID' and 'Phase ID'
these are automatically
specified as batch and
phase ID.

SIMCA import
103
Class ID specification
Specifying the Class ID at import defines classes in the default workset.
Only one column can be defined as Class ID and it is only available for regular
projects.
Use one of the following methods to open the Class ID dialog:
- Mark the column and then on the Home tab, in the Identifiers group, click
Class ID.
- Mark the column, click the column arrow, and then click Class ID.

In the Split ID dialog, enter the start and length of the characters specifying the
classes, or click OK to use the entire string. Length '-1' denotes 'until the end of the
string'.

Class pane
After specifying a Class ID, the Class window is updated accordingly.
Use the Class pane to Rename, Merge, Delete, and switch the order of the classes or
excluding variables in certain classes. For more about these actions, see the Batch &
Phase pane phase part subsection later in this chapter.
Note: Deleting classes here will omit them in the import while excluding
variables only exclude them in the default workset.

User Guide to SIMCA
104
Batch and Phase ID specification
To specify the batch and phase IDs:
1. Mark the column and click Batch ID or Phase ID.
2. Specify the position of the IDs by entering start character position and
length.

If the Batch or Phase IDs span several columns, use the Merge Column button in the
Edit tab to concatenate them.
Specifying a Batch ID updates the content of the Batch & Phase pane. This pane is
parted in two when Phase IDs have been specified: a phase part and a batch part.
Downsizing using conditional exclude
When the file is large and contains many batches, conditional exclude can be used to
downsize the number of batches.
There are 2 types of conditional exclude available:
Cond. exclude type Action by the user in the dialog
Excluding all batches, all
phases, or the selected
phase in each batch with
less than / more than the
user entered number of
observations.

1. Select the first Exclude option.
2. Select all batches, all phases, or one of the
phases.
3. Select with less than or more than.
4. Enter the number of observations that defines
the limit.
SIMCA import
105
Cond. exclude type Action by the user in the dialog
Excluding a user entered
percentage of all batches at
random / the longest / the
shortest / ordered.

1. Select the second Exclude option.
2. Enter the percentage to use in the field.
3. Select how to delete the entered percentage of
batches by clicking at random/the
longest/the shortest or ordered.
Excluding/including rows or columns
To exclude or include a column or row in the import wizard spreadsheet: mark it and
click the Exclude or Include-button.
Excluded rows and columns are colored gray.

Specifying data properties
For a regular project (non-batch), with no qualitative variables, it is necessary to
include all variables (set them as x) but not necessary to specify y-variables.
Specifying y-variables defines the default workset.
With batch data, the x variable, y variables (time or maturity), and variables pertaining
to the whole batch, such as batch conditions, X or Y (quality of the batch) all need to
be specified in the SIMCA import.
All about specifying data
The table lists the Variable Types available in the SIMCA import, their description,
the objective with specifying the type of variable, and the spreadsheet coloring after
specification.
Specifying variable types is done as follows:
1. Mark the columns.
2. Click the relevant button in the Variable Types group:
- Quantitative
- Qualitative
- Date/Time
- Batch Conditions - requires that the column is also Quantitative
or Qualitative.
- X Variables
User Guide to SIMCA
106
- Y Variables
Note: All data columns have to be specified as X Variables or Y Variables in
addition to Quantitative/Qualitative or Date/Time.

Button Description and objective Coloring
Quantitative Continuous variables.
Imports a quantitative variable.
The variable is
colored white.
Qualitative Discrete variables that are split into
dummy variables during the
calculations.
Imports a qualitative variable.
The variable is
colored turquoise.

Date/Time
Variable
Specifies the selected variables
according to the specified time format.
For more, see the Formatting
variable as Date/Time section later in
this chapter.
Date/time variables can be displayed
on the x-axis showing time in the
specified format.
The variable is
colored orange.
Batch
conditions

Batch condition variables can be
imported with the batch evolution
dataset or as a separate dataset
arranged as a batch level dataset.
Makes batch conditions available for
modeling in batch level models.
The variable is
colored light grey.
X Variable All numerical variables, except the first
column, are by default assumed to be
x-variables.
The x-variables are the explanatory
variables.
Includes the variable in the import.
The variable is
colored white.
Y Variable Defines the default workset as PLS.
The y-variables are the predictor
variables.
Specifies the default workset.
The variable is
colored light brown.
Specifying time or maturity
Batch evolution models, BEM, require a y-variable, time or maturity. Neglecting to
specifying a y-variable results in that $Time, automatically created by SIMCA, is used
as y. $Time = relative local batch or phase time (i.e. the time restarts at the beginning
of each batch and phase) with a sampling interval of 1.
SIMCA import
107
Formatting variable as Date/Time
A variable specified as Date/Time at import can be used:
- As a variable in the model. The date and time variable is then transformed
into a regular value (float).
- As x-axis in plots. The Date/Time variable is by default displayed in line
plots, for instance the Score Line Plot and DModX.
Specifying the Date/Time variable
The autoformatting tries to recognize date/time variables and specify them as such. If
the variable is not by default specified as date/time, follow the steps in the table:
Step Description Illustration
1. Select the variable in the
spreadsheet by clicking its
column number.
With more than one variable in
the same format, mark them all.

2. Specify the variable type by
clicking Date/Time. If the
variable is specified as X it is by
default excluded from the
default workset, if specified as
Y, it is default included.

3. In the message box, the
guessed date/time format is
displayed if one is found. If
SIMCA cannot find an
appropriate date/time format the
dialog in 4a is displayed.

User Guide to SIMCA
108
Step Description Illustration
4a

Clicking No opens the Specify
Date Format dialog.
Enter the format for the
date/time variable and click OK
to reopen the dialog in step 3.
Click the Show Details button
to view the terminology, also
displayed in the Date/Time
formatting terminology
subsection later in this section.

4b Clicking Yes opens the Specify
Time Unit dialog.
In this dialog the format for
storage is selected. The unit
selected here is the smallest
unit that is guaranteed to be
displayed correctly.
Worth noticing here is that:
- Selecting to store the
variable in seconds limits
the number of days that
data can be imported for
and the accuracy of the
saved value can be
guaranteed.
- Selecting to store the data
in a larger unit than the
imported format will result
in rounding errors when
displaying the date/time
variable in the more
detailed format.

5 Clicking OK saves the format to
use for parsing the variable.

For reformatting after importing the dataset, see the Date/Time configuration in
dataset properties subsection in the Data chapter.
Date/Time formatting terminology
The following format details are available when clicking Show Details:
Column and row menus
Formatting rows and columns, excluding, and merging can be done by marking one or
more columns/rows and clicking the column or row arrow.
A menu opens with the following commands for columns: Primary ID, Secondary
ID, Class ID (regular projects only), Batch ID (batch projects only), Phase ID (batch),
Batch & Phase ID (batch), Quantitative Data, Qualitative Data, Date/Time
Variable, Batch Condition (batch), X Variable, Y Variable, and Exclude Column.
SIMCA import
109
For rows the following commands are available: Primary ID, Secondary ID, Exclude
Row, and Include Row.
All of these commands are available from the buttons on the Home tab and are not
described here.

Importing batch conditions
To import initial and final conditions, in SIMCA named batch conditions, and use
them in a BLM, click File | Import dataset. After import the new dataset can be
selected in the Workset dialog when the Create a batch level model check box is
selected.
When importing a dataset in a batch project and each batch only has one row, the data
are assumed to be batch conditions.
Note: Beware of how you name your batch condition variables to not confuse
them with the score names generated from the BEM. That is, do not use names
such as t(1)_2_20.
Finish and create project
After opening all datasets in the SIMCA import, click Finish to create the SIMCA
project.

In the dialog, select destination and enter project name.
User Guide to SIMCA
110
Edit tab
On the Edit tab, there are five groups: Indexes, Rows & Columns, Merge,
Formatting, and Sort. This section describes the features available in these groups.
Creating observation or variable index
To create an index column or row, on the Edit tab, in the Indexes group, click
Column or Row depending on which index you want to create.

Rows and Columns
The standard editing commands Insert Row, Insert Column, Delete (Clear, Entire
Rows, Entire Columns etc.) and Shift are available in the Rows & Columns group in
the Edit tab.
Additionally Transpose and Downsize, described next in this section, are available

Transpose
To transpose the dataset, on the Edit tab, in the Rows & Columns group, click
Transpose.
Transposing the dataset removes all formatting.

Downsizing the dataset
To use a subset of the imported dataset:
1. On the Edit tab, in the Rows & Columns group, click Downsize.
2. In the Start after xx observations field, enter the starting observation. All
observations, before and including this observation, remain included.
3. In the Keep every 'xx' observations field, enter the interval to use when
downsizing.
SIMCA import
111

Downsizing is not available here for batch projects. For batch projects, downsizing is
available as Conditional exclude in the Batches page in the SIMCA import. For more,
see the Batch & Phase pane batch part subsection later in this chapter.
Merge
Split Column
To split a column in two, mark the column and click the Split Column button in the
Merge group.
Splitting a column is useful when you have different information in the same column
and would prefer them in separate columns.
Merging columns
To merge columns:
1. Mark the columns.
2. On the Edit tab, click Merge Columns.
3. In the Merge Columns dialog, select the order in which the columns should
be merged by marking and clicking the up and down arrows in the dialog.
4. If the columns should be merged with a separator, click Separate the
content in each column using and enter a separator. If the columns should
be merged without separator, select Concatenate the columns without a
separator.
5. Click OK and the columns are merged and the original ones excluded.

User Guide to SIMCA
112
Merging Spreadsheets
You can import one or more files using the SIMCA import. When in SIMCA, datasets
selected in the Workset dialog are merged by primary ID behind the scenes and
merging is not necessary.
If you want to merge data before creating the project, on the Edit tab, in the Merge
group, click Merge Spreadsheets.
The table lists a description of the steps.
Step Objective and explanation Action
1. Importing the datasets. Click File | New
Spreadsheet and import the
datasets to merge in SIMCA
import.
2. Initiating the merge of two or more files
in one single dataset.
On the Edit tab, in the Merge
group, click Merge
Spreadsheets.
3. Selecting the files to merge. In the Merge dialog, clear
spreadsheet check boxes so
that only the files to merge
are selected.
4. Sorting the files. Move the files using the
arrows leaving the files in the
desired order.
5. Merging matching identifiers/index.
When merging and matching by identifier,
that identifier needs to be selected in the
Merge By box.
Selecting to merge by Index merges on
row/column index.
In the Merge By box, select
Index or Primary ID. If no
primary ID has been
specified, only Index is
available.
6. Selecting direction to merge. This
option determines how the merging is
done, positioning the datasets next to
each other in Side by Side or above and
below in Top/Bottom.
In the Direction box, select
in which direction to merge,
Side by Side or
Top/Bottom.
7. Selecting the form of the merged data.
Destination: The destination file (first
one in the dialog) specifies which
observations/variables are included.
Intersection: The observations/variables
common to all selected datasets are
included.
Union: All observations/variables present
in any of the datasets are included.
In the Result box select
whether the result is
variables/observations
according to the Destination
file, the Intersection or the
Union.
8. Completing the merge. Click OK to merge the files
as specified deleting the
original spreadsheets.

SIMCA import
113
Apply Formatting
Formatting specified for a file previously imported can be applied to the current
spreadsheet by clicking the Apply Formatting button and selecting the spreadsheet
name. The files listed are those with matching variable and observation IDs.
To autoformat (autocolor) the dataset, click Apply Formatting | Autoformat.
Clear formatting by clicking the Clear Formatting button in the same group.
The formatting features are available in the Formatting group on the Edit tab.
Note: A row, with slash </> as the leftmost character of the leftmost column
of that row, is automatically marked as excluded by SIMCA. A column with
slash </> as the leftmost character in the top cell is automatically marked as
excluded by SIMCA.
Sort
Sort Ascending and Sort Descending are available when marking one column only
(click the column header). Clicking the button results in that the entire dataset is sorted
according to the marked variable.

View tab
The information panes Issues, Audit Trail, Variables, Observations, Batch & Phase
and Summary can be shown or hidden from the View tab.
The Missing Value Map is also available here. See the Displaying the Missing Value
Map section.
Arranging windows cascaded or tiled is available by clicking Cascade, Tile
Horizontally and Tile Vertically. All open spreadsheets are arranged accordingly.

Issues pane in SIMCA import
The Issues pane lists problems in individual sheets as well as between sheets. When
there is an issue that the SIMCA import can resolve for you the Select Action column
displays what the action will be when Resolve All is clicked. Hover over the issue to
display more information in a tooltip.
As long as there are unsolved issues, clicking Finish will not let you create a project.
User Guide to SIMCA
114
Issues and actions in the Issues pane
The table lists some common issues, the reason for the issue, possible ways to resolve
and the result from the selected resolution.
Note: Hover over the issue for more info.
Issue Description
Actions
available
Result after
Resolve All
No primary
Obs ID
No primary
Var ID
The ID has not been
specified.
None
Clicking Resolve All
or Finish both add
unique primary IDs.
Missing
observation
ID
Missing
variable ID
Missing batch
name
Missing
phase name
There are missing
values in the
column/row specified as
primary/batch/phase ID
rename
auto rename
exclude
row/column
rename - opens a
dialog allowing you to
enter an ID.
auto rename - adds a
number to create
unique IDs.
exclude row -
excludes the problem
rows/columns.
Primary
observation
ID is not
unique
Primary
variable ID is
not unique
There is at least one
duplicated primary ID.
rename
auto
generate
exclude
row/column
rename - opens a
dialog allowing you to
enter the new ID for
the duplicate.
auto generate -
creates new
observation/variable
IDs for all.
exclude row -
excludes the
rows/columns with
duplicated IDs.
No batch ID
No batch ID was
specified for the batch
dataset.
auto
generate
auto generate is
available when a
phase ID has been
specified - creates
batch ID using the
phase ID.
No phase ID
No phase ID was
specified for the batch
dataset. Displayed
when importing a
dataset to a batch
project with phases

This has to be
resolved by specifying
a phase ID.
SIMCA import
115
Issue Description
Actions
available
Result after
Resolve All
Batch
condition
variable lacks
primary
variable ID
A variable specified as
batch condition lacks
primary variable ID.
rename
auto rename
exclude row
rename - opens a
dialog allowing you to
enter an ID.
auto rename -
suggests an ID.
exclude row -
excludes the problem
rows.
More than
one dataset
have the
name ...
Each dataset in a
project needs to have a
unique name. All
datasets are default
named by the file name.
rename
Note here
that clicking
the Rename-
button to the
left of the
issue allows
you to
specify the
new dataset
name
rename - adds a
number making the
dataset name unique.
Unrecognized
qualitative
setting in
variable
For a project with one
or more models, no new
settings in a qualitative
variable can be
imported. To import a
new setting, delete all
models, including
unfitted models before
import.
treat as
missing
exclude row
exclude
column
treat as missing -
replaces the new
setting with missing.
exclude row -
excludes all rows with
the new setting.
exclude column -
excludes the column.
Variable type
mismatch
A variable must be of
the same type in all
datasets in a project.
The Variable type
mismatch warning
appears when the same
primary variable ID was
specified for different
variable types, e.g.
quantitative/qualitative/d
ate time.

Maturity is
not specified.
A variable
called $Time
will be
generated
No y-variable has been
specified.
None
Clicking Resolve All
or Finish both add
the $Time variable.
User Guide to SIMCA
116
Issue Description
Actions
available
Result after
Resolve All
Non-
continuous
batch IDs
All data for a given
batch must be
continuous.
auto rename
(default)
exclude row
sort
auto rename - adds a
number to the batch
IDs that are perceived
to be out of order.
exclude row -
excludes the first row
that appears out of
order and continues
until the batch IDs are
in order.
sort - sorts the batch
IDs
Non-
continuous
phase IDs
All observations in a
phase must be
consecutive within
every batch.
manual
exclude
(default)
split batch
exclude last
in batch
exclude first
in batch
manual exclude -
results in nothing
when clicking
Resolve All. You
have to decide what
to do yourself.
split batch - adds
number to all batch
IDs following the
observation with the
problem in phase ID.
exclude last in batch -
excludes all rows
following the problem
phase ID in that
phase.
exclude first in batch -
excludes all rows
preceding the row
with the problem
phase ID.
Existing
batch
variable
combination
Datasets in the same
project may only
contain the same
batches OR the same
variables, not both.
rename
batches
(default)
manual
merge
rename batches -
adds a number
making the batch IDs
unique.
manual merge -
results in nothing
when clicking
Resolve All. You
have to decide what
to do yourself.
SIMCA import
117
Issue Description
Actions
available
Result after
Resolve All
Existing Var.
Obs.
combination
Datasets in the same
project may only
contain observations
OR variables with the
same primary IDs.
rename
observations
(default)
rename
variables
exclude row
exclude
column
rename observations
- adds a number to
the primary IDs that
are found matching,
making them unique.
rename variables -
adds a number to the
primary IDs that are
found matching,
making them unique.
exclude row -
excludes the rows
with matching IDs.
exclude column -
excludes the columns
with matching IDs.
Conflicting
batch ID
Conflicting
phase name
Conflicting
class name
Identical primary
observation IDs may
not be associated with
different
batch/phase/class IDs.
auto rename
exclude row
auto rename - adds a
number making the
IDs unique.
Phases out of
order
The phases need to be
in the same order for all
batches.
reorder
phases

reorder phases -
orders the phases in
the order of the
phases in the Batch
& Phase pane.
Phases not in
the same
order in all
datasets
The phases need to be
in the same order for all
batches in all datasets.
manually
reorder
phases
The phases have to
be reordered. You
can do this easily for
each dataset using
the Batch & Phase
pane.
Invalid value
String that is not in a
row or column with
identifiers or in a
column defined as
qualitative.
treat as
missing
exclude row
exclude
column
treat as missing -
replaces the content
of the cell with
missing.
exclude row -
excludes all rows with
the invalid values.
exclude column -
excludes the columns
with the invalid
values.
User Guide to SIMCA
118
Issue Description
Actions
available
Result after
Resolve All
Variable IDs
cannot
contain any
of the
following
characters: $
Observation
IDs cannot
contain any
of the
following
characters: $
$ is reserved for SIMCA
created vectors.
rename
auto rename
exclude
column/row
rename - opens a
dialog allowing you to
enter an ID.
auto rename - adds a
number to create
unique IDs.
exclude row/column -
excludes the problem
rows/columns.
Incorrect
date/time
variable
settings
The content of the cell
cannot be formatted as
date/time.
treat as
missing
exclude row
exclude
column
treat as missing -
replaces the content
of the cell with
missing.
exclude row -
excludes all rows with
the invalid values.
exclude column -
excludes the columns
with the invalid
values.
Unexpected
phase ID
The first dataset in the
project does not have
phases, then other
datasets may not either.
exclude
exclude - excludes
the column.
Non printable
character
ASCII characters 1 - 31.
exclude
rename
exclude - excludes
the row/column.
rename - opens the
rename dialog.
Audit trail
The audit trail needs to be turned on before importing the first dataset to hold all
information, see the Options subsection earlier in this section.
Note: The Audit Trail is empty until clicking Finish when the Audit Trail
starts logging information later displayed in the Audit Trail in SIMCA.
For more, see the Audit Trail section in the View chapter.
Variables pane
The Variables pane lists the number of variables currently included, primary variable
IDs of the currently included variables, variable type and % missing.
Using the shortcut menu you can reach the same commands as in the Column and row
arrow menus, for more see the Column and row arrow menus subsection earlier in
this section.
SIMCA import
119
Observations pane
The Observations pane lists number of observations currently included, primary
observation IDs of the currently included observations, and % missing.
Using the shortcut menu you can reach the same commands as in the Column and row
arrow menus, for more see the Column and row arrow menus subsection earlier in
this section.
Class pane
After specifying a Class ID, the Class window is updated accordingly.
Use the Class pane to Rename, Merge, Delete, and switch the order of the classes or
excluding variables in certain classes. For more about these actions, see the Batch &
Phase pane phase part subsection later in this chapter.
Note: Deleting classes here will omit them in the import while excluding
variables only exclude them in the default workset.

Batch & Phase pane phase part
With a batch project with phases, the phase part of the Batch & Phase pane lists:
- Each phase.
- The number of batches containing that phase (batches do not have to include
all phases).
- The number of variables in that phase (phases may have different variables).
- The roles of the variables.
User Guide to SIMCA
120

Note: When moving/deleting/rearranging phases in the Batch & Phase pane,
nothing happens in the dataset spreadsheet until the dataset is imported to
SIMCA.
Excluding variables
If the phases have different variables you can exclude variables in the following
manner:
1. Expand the node in each phase to list the variables.
2. Mark the variables to exclude.
3. Click the Exclude-button.
This excludes the variables in the default workset. You can also configure the
variables for each phase, after import, in the Workset dialog.
Delete phases
Delete a phase in a certain batch by marking and clicking the Exclude-button.

Note: Deleting phases here will omit them in the import while excluding
variables only exclude them in the default workset.
Switching the order of the phases
If phases are imported in the incorrect order, mark the phase and click the up or down
arrow in the dialog.

SIMCA import
121
Rearranging phases automatically
Let SIMCA rearrange the phases automatically by clicking the Rearrange phases
button. The phases will then be arranged in the order in the Batch & Phase pane.

Merging phases
To merge phases, mark two or more phases and click the Merge button.

If you want to merge phases positioned in two different datasets, you first need to
merge the datasets. See the Merging Spreadsheets subsection for more.
Note: You can only merge consecutive phases.
Assigning different time or maturity variable and renaming
phases
If you want to assign different time/maturity variables to the different phases and/or
rename a phase, mark a phase, click Configure (last button in the Batch & Phase
pane), and select the desired time/maturity from the Time/Maturity variable-box.

Batch & Phase pane batch part
The Batch & Phase pane Batches section displays batches, the phases present in each
batch, and the number of observations in each phase for each batch.

Expand the node to see all the phases in a batch with the number of observations in
each phase.
User Guide to SIMCA
122
Excluding batches or phases
Exclude batches or a phase in a certain batch by marking and clicking the Exclude-
button .
Summary pane
The summary pane lists the following information for the current dataset:
- Source - file name and path when data was imported from file.
- Number of variables.
- Number of observations.
- Missing values.
- Number of classes.
- Number of phases.
- Number of batches.
Displaying the Missing Value Map
To display an overview of a dataset with respect to missing values:
- On the Data tab, in the Summary group, click Missing Value Map and then
the desired dataset in SIMCA.
- In SIMCA import, on the View tab in the Missing Values group, click the
Missing Value Map.
Missing values are colored while data present are white.


123
Home
Introduction
The Home tab holds the commands most commonly used when creating models and
evaluating them. On the Home tab you can:
- View the imported datasets - Dataset spreadsheet
- View statistics and create a new model using the commands in the Workset
group.
- Fit the model - using the Fit model group.
- View diagnostic and interpretation plots: Summary of Fit, Scores,
Loadings, Hotellings T2Range, Distance to Model, Observed vs.
Predicted, Coefficients and VIP.

Dataset spreadsheet
To open a dataset in a spreadsheet:
- Click Dataset on the Home tab.
- If there are more than one dataset, select the desired dataset by clicking the
arrow in the lower part of the Dataset-button and clicking it. The datasets are
displayed in the order they were imported.

Shortcut menu in the dataset spreadsheet
Open the shortcut context sensitive menu by right-clicking the spreadsheet.
User Guide to SIMCA
124
The following commands can be available: Undo, Cut, Copy, Paste, Delete, Create |
List/Plot, Create | XObs/YObs, Lock the First xx Columns/Rows, Add
Observations/Variables, Sort Ascending/Descending, Add Secondary
Observation/Variable ID, Quick Info | Variables/Observations, Zoom Out/In,
Missing Value Map, Trimming Overview, Format Date/Time, Save List As, Print,
and Properties.
All commands are available from the ribbon and described in the chapter for the
relevant tab. The following commands are dataset specific and not described
elsewhere:
1. Adding observations and variables.
2. Adding secondary IDs.

Adding observations and variables
Adding a specified number of observations or variables in the dataset spreadsheet is
available from the shortcut menu.
To add observations or variables:
1. Right-click the dataset spreadsheet and click Add Observations or Add
Variables.
2. In the Add Observations or Add Variables dialog, type the number of
observations or variables to add.
3. Click OK and note that empty rows or columns are added at the bottom or to
the right in the dataset spreadsheet.
4. Paste or type the values for the new observations or variables.

Home
125
Note: Adding variables as a function of other variables or model results is
available using Generate Variables. Generate Variables is available on the
Data tab, in the Modify Dataset group.
For information about adding variables as a function of other variables or model
results, see the Generate Variables section later in this chapter.
Adding secondary IDs
Secondary IDs can be added directly in the dataset spreadsheet, by right-clicking and
selecting Add Secondary Observation ID or Add Secondary Variable ID. Empty
identifiers are added. Paste or type their content.
Workset
This section describes the features in the Workset group on the Home tab.
The workset is all or a subset of the selected datasets, with specifications for
scaling, transformations, variable roles, lags and expansions of variables. Observations
can be grouped in classes for SIMCA classification or discriminant analysis. In batch
projects with phases the observations are grouped in phases.
When exiting the workset dialog, SIMCA creates an active model, linked to that
workset. A workset is associated with every model.
Content
In the Workset group the following commands are available:
1. Statistics | Workset Statistics displays the selected statistics for the current
workset; Correlation Matrix displays the correlations for the variables
included in the current workset.
2. New opens the workset dialog with the settings of the default workset.
3. New As | Mx opens the workset dialog using the settings of the selected
model.
4. Edit | Mx opens workset dialog of the selected model.
5. Delete | Mx deletes the selected model and workset.
6. Change Model Type lets the user select among the available model types.
7. The Model Options dialog box launcher opens the Model Options dialog
for the current workset, for editing and viewing.

User Guide to SIMCA
126
Workset Statistics
To display descriptive statistics for selected variables in the current workset, click
Statistics | Workset Statistics.
Note: When marking a CM, BEM or DA-model, the statistics are calculated
over all classes/phases.
The default Workset Statistics list displays N (number of observations), Missing
values (%), Mean, and Standard deviation for the variables in the workset.
To view more than the default statistics, open Properties and select what to display.
The Properties dialog is described next.
Workset Statistics Properties
The Workset Statistics Properties dialog has two sections: Display and Use only.

Display in Workset Statistics
In the Display section, the statistics are listed.
The available statistics are: N, Missing value (%), Min, Max, Min/Max, Mean,
Median, Standard deviation, Std. dev/Mean, Skewness, Skewness test, and
Kurtosis.
By default the N, Missing value (%), Mean, and Standard deviation check boxes are
selected.
To display other or more statistics, select the desired check boxes.
Use only in Workset Statistics
The statistics are computed on the included observations in the active model, for the
variables selected in the Use only list. The statistics are displayed for the transformed
and trimmed variables when specified in the workset.
By default all variables are included.
To view the statistics of a selection of variables, mark the variables in the Use only list.
Home
127
Correlation Matrix
The correlation matrix is a spreadsheet that shows the pair-wise correlation between all
variables (X and Y) in the current workset, scaled and transformed as the workset.
Each variable is displayed on one row and one column in the correlation matrix, and
the correlation between two variables is shown in the cell where the two variables
intersect. By double-clicking a cell the corresponding scatter plot of the raw data is
created. The value of the correlation coefficient represents the extent of the linear
association between the two terms. The value of the correlation coefficient ranges from
-1 to 1. When the correlation coefficient is close to zero there is no linear relationship
between the terms.
Note: When marking a CM, BEM or DA-model, the correlations are
calculated over all classes/phases.
To display the Correlation Matrix for the current model, click Statistics |
Correlation Matrix.


Coloring
SIMCA uses a coloring scheme in ten levels (from dark color to white) to assist in the
interpretation of the correlation matrix. The darker the color, the higher the absolute
correlation.
Limitation
By default the correlation matrix is restricted to 32 variables, but this limit can be
changed in the SIMCA Options dialog.

User Guide to SIMCA
128
New workset
To create a new workset from the settings of the default workset, click the New-
button. The Workset dialog opens.
For more about the Workset dialog, see the Workset dialog section later in this
chapter.
Default workset
When exiting the SIMCA import the default workset is created according to
specifications during the import. For batch projects all datasets imported when creating
the project are by default included while only the first dataset is included when
creating a regular project.
The properties that can be specified for the default workset are:
- Selected datasets.
- Variables as X, Y, or excluded.
- Variable scaling but only for the X respective Y variables as groups, not for
individual variables.
- Observation class belonging.
Note: No lags, expansions, or transformations can be saved in the default
workset.
Specifying the default workset
The default workset can be specified as follows:
- Datasets included: Select the desired datasets in the Select Data page in the
Workset dialog and click Save as Default Workset.
- Y-variables: Specify variables as Y in:
o the SIMCA import or
o the Workset dialog, Variables page, then click Save as Default
Workset. Here the configuration of a batch project y-variable can
also be saved.
- Exclude variables: Specifying variables as Excluded in the Workset dialog,
Variables page.
- Scaling: Specifying X and Y variable scaling in File | Project Options, tab
Fit, in the Default Scaling section.
- Classes: Specifying class belonging
o by specifying Class ID in the SIMCA import or
o in the Dataset Properties, tab Observations.
Home
129
New workset as model
To create a new workset copying the settings from a selected model, click New As and
select the desired model.
New as model with class models
With class models, the New As menu holds both wrapper CM and class models.

Clicking New As |
- CMxx opens the workset dialog with the variables of all classes and the
same specification as model CMxx.
- xx, opens the workset dialog with only the observations and variables of the
selected class and the same specifications as model xx. Any change to
variables, scaling or transformation of variables affects only that class
(model). Clicking OK creates an unfitted model for the selected class in a
new CMxx wrapper.
New as model for batch projects with phases
With batch project with phase models, the New As menu holds both wrapper BEM and
phase models.

User Guide to SIMCA
130
Clicking New As |
- BEMxx opens the workset dialog with the variables of all phases (classes)
and the same specification as model BEMxx.
- Mxx, opens the workset dialog with only the observations and variables of
the selected phase and the same specifications as model xx. Any change to
variables, scaling or transformation of variables affects only that phase
(model). Clicking OK creates an unfitted model for the selected phase in a
new wrapper BEMxx.
Editing the workset
To edit the workset of a selected model, click the Edit-button and select the desired
workset. If the model is fitted, it is replaced by the edited unfitted model.
Editing a model is particularly useful when a transformation or scaling of a variable is
necessary in one phase or one class only.
Editing models with class models
With class models, the Edit menu holds both wrapper CM and class models.
Clicking Edit |
- CMxx opens the workset dialog with the variables of all classes and the
specification as model CMxx.
- xx, opens the workset dialog with only the observations and variables of the
selected class (model). Any change to variables, scaling or transformation of
variables affects only that class. Clicking OK replaces the previous model
with an unfitted model.
Editing models for batch project with phases
For a batch project with phase models, the Edit menu holds both wrapper BEM and
phase models.
Clicking Edit |
- BEMxx opens the workset dialog with the variables of all phases (classes)
and the specification as model BEMxx.
- Mxx, opens the workset dialog with only the observations and variables of
the selected phase (model). Any change to variables, scaling or
transformation of variables affects only that model. Clicking OK replaces the
previous model with an unfitted model.
Home
131
Deleting the workset
To delete a workset, click the Delete-button, and select the workset. Both the workset
and the model associated with it will be deleted.

Workset dialog
After selecting to create a new workset, the Workset dialog opens with the current
observations and variables and their attributes.
Regular project Batch project

The workset is organized in tabs, each holding a page. Clicking through the tabs opens
the desired pages to change the attributes of the observations or variables. The
available tabs are Select Data, Overview, Variables, Observations, Batch,
Transform, Lag, Expand, Scale, and Spreadsheet. A description of each page
follows later in this section.
Note: The list on each of the Workset dialog pages can be copied by selecting
the items to copy and then pressing CTRL+C.
The first subsections describe features available on all or close to all pages. These
features are:
- Model type box - described in the Model types in the Workset dialog
subsection next. Unavailable for batch evolution models, BEM.
- Find feature - described in the Find feature in workset dialog subsection.
Available on all pages but the Select Data, Overview and Spreadsheet
pages.
User Guide to SIMCA
132
- As model box - described in the As model subsection. Available on all pages
but the Batch, Scale, and Spreadsheet pages.
- Create partial models check box - described in the Partial models for
batch level subsection. Available for batch level models only.
The workset wizard, available for regular project, is described in the Simple mode
workset wizard section later in this chapter.
Note: Pressing ENTER or clicking OK, at any time, exits the workset. To
continue defining the workset dialog, click another tab.

Note: When marking a variable in one page, it remains marked in all pages
when clicking another tab.
Model type in the Workset dialog
The default model type depends on the workset specifications.
Click the Model type box to change the model type.
Available model types in the Workset dialog
The possible model types, in the Model type box, are: O2PLS, O2PLS-Class,
O2PLS-DA, O2PLS-Hierarchical, OPLS, OPLS-Class, OPLS-DA, OPLS-
Hierarchical, PCA-Class, PCA-Hierarchical, PCA-X, PCA-X&Y, PCA-Y, PLS,
PLS-Class, PLS-Hierarchical, PLS-DA and PLS-Tree.

Default model type
The default model type is set as follows:
- PLS if there are continuous y-variables.
- PLS-Class if there are continuous y-variables and observation classes
defined.
- PLS-DA if there is one qualitative y-variable.
- PLS-DA if there are no y-variables and 6 or less classes.
- PCA-Class if there are no y-variables and more than 6 classes.
- PCA if there are no y-variables and no classes.
- PLS-Hierarchical if variables have been assigned to blocks and there are y-
variables.
- PCA-Hierarchical if variables have been assigned to blocks and there are no
y-variables.
Note: The previously selected model type is default in new worksets with the
same workset specification.
Home
133
Find feature in workset dialog
The Find feature is available in all of the pages but the Select Data, Overview and
Spreadsheet pages of the Workset dialog, although with some variations.
There are two types of Find available from the arrow: searching independent of the
Find field and the searching according to the entry in the Find field.

Select All, Complement Selection, Select
The first find options are independent of the Find field:
- Select All (CTRL+A) marks all items in the list.
- Complement Selection (CTRL+I) marks all unmarked items and deselects
the previously marked.
- Select opens a dialog with context sensitive options. Available in the
Variables and Observations pages.
For observations:

For variables

User Guide to SIMCA
134
Find in column
The result from using Find starting with 'Find' depends on what is entered in the Find
field.
In the first section all IDs are displayed, and for the Transform and Scale pages also
the displayed statistics. Default is to Find in Primary ID Column.
The second section defines how the search is done: Find Beginning With, Find
Containing, or Find Exact. Default is Find Beginning With.
Note: In the Transform and Scale pages the Find utility can search in the
statistics columns too.
Find field
When typing in the Find field, the search is done according to the current settings,
viewed by clicking the arrow.
Wild card symbols '?', and '*' are allowed in specifying observations or variables
names. For example ?LH* selects observations or variables with names such as
SLH2 or QLHSW, etc.
To increase or decrease the time before the search starts after typing in the Find field,
change the delay time in SIMCA Options. For more, see the SIMCA Options
subsection in the File chapter.
Note: The Find utility in SIMCA is cASe inSensiTiVE.
As model
The As model box is available on the Select Data, Overview, Variables,
Observations, Transform, Lag, and Expand pages. They allow selection of the same
settings as another model according to the table:
Page Settings copied
Select Data Dataset selection.
Overview Variable settings according to Variables, Transform, Lag,
and Expand below.
Observation settings according to Observations below.
Variables Variables and roles as the selected model.
Observations Observations with the same class structure as specified in
the selected model.
Transform The same transformations for the same variables as the
selected model.
Lag Lag structure of the selected model.
Expand Expanded terms as in the selected model.

Home
135
Partial models for batch level
For batch level datasets, partial models can be generated to predict the quality of the
batch, or classify the batch before completion, by selecting the Create partial
models check box found at the bottom of the Workset dialog.

There are two types of partial models, one according to phases and one according to
completion:
Steps Description with and without phases
1. Select
Create
partial
models
check box
in the
Workset
dialog.
With phases
The check box just becomes selected.
Without phases
SIMCA opens the following dialog:

In the Number field, type the number of partial models to create after
Number. The variables are then parted according to maturity. Click OK.
2. Click
OK to exit
the
Workset
dialog.
With phases
Automatically fits the models. The partial models are built sequentially
from the phases.
Without phases
Automatically fits the models. The partial models are built from percent
completion.
3. View
the result.
With phases

Without phases

User Guide to SIMCA
136
Select Data page
The Select Data page is available when there are more than one dataset available in the
project.
The Select Data page lists all datasets in the project for regular projects. For batch all
batch evolution datasets are listed when the Create a batch level model check box is
cleared and all batch level datasets are listed when the Create a batch level model
check box is selected.
When selecting more than one dataset, the datasets are merged by primary ID in the
workset. The Primary ID match percentage is updated to indicate how well the
primary IDs in the selected datasets correspond.
To see how the merged datasets will be arranged during the calculations, click the
Spreadsheet tab. Select workset as raw data in the View box if you want to view the
data in original units.
Overview page
The Overview page displays the variables and observation lists with their respective
attributes.
Available from the Overview page are:
- The variable list displaying the current role of each variable, X or Y, primary
variable ID, transformations, expansions, scaling, scaling block, and lags.
- The observation list displaying primary observation ID and when available,
Class IDs, Batch IDs, and Phase IDs.
- Shortcut menu both in the variable and observation lists with the same
commands as in the Variables and Observations pages. For more, see the
Variables page and Observations page subsections later in this chapter.
- Missing value tolerance to specify accepted percentage of missing in
variables and observations, and checking missing values.
- As model for assigning X and Y etc. as another model. See the As model
subsection, previously in this chapter for more.
- Use Simple Mode - see the Simple mode workset wizard subsection later
in this chapter.
Variable list - Overview page
The top part of the Overview page displays summarized variable information:
- Variables: Number of included variables in the workset, X and Y.
- Expanded terms: Number of expanded terms in Expand page.
- Lags: Number of lags defined on the Lag page.

Home
137
In the variable list, the X variables are listed first and Y variables last, and with the
following information:
1. Variable role displayed as an X or a Y to left of the variable ID.
2. Primary Variable ID under Primary ID.
3. Defined transformations under Transformation.
4. E under Expansion indicates that the variable is expanded with
cross, square, or cubic terms.
5. Type of scaling under Scaling.
6. Scaling block number, if the variable is part of block scaling under S. Block.
7. L indicating that the variable is lagged under Lag.
Lags and expansions appear at the bottom of the list.
Observation list - Overview page
Above the observation list the number of observations included is displayed.
The observation list holds the observation primary ID and class belonging, or for batch
projects batch, and when available, phase IDs.


1. Primary Observation ID under Primary ID.
2. Class of the observations under Class, if the observations are grouped in
classes.
3. Batch and phase IDs under Batch and Phase for batch projects.
Missing value tolerance
For the results of model fitting to be reliable, the workset should contain less than 50%
missing values in observations or variables. This is the SIMCA default threshold for
missing values for the workset and for predictions.
The Missing value tolerance section is positioned at the bottom of the Overview
page. It contains the default missing value tolerance for Variables and Observations
and a Check Missing Values button.
Missing values percentage exceeding tolerance
When clicking the Check Missing Values button, or OK, the observations and
variables missing values percentage is checked versus the tolerance limits. When the
missing values percentage of a variable or observation exceeds the specified tolerance
limit, SIMCA prompts for including or excluding the variable or observation.
When no variables or observations exceed the missing value tolerance, the following
message is displayed after clicking the Check Missing Values button:

User Guide to SIMCA
138
Changing the missing value tolerance limit
Change the tolerance limit displayed in the window of the workset by typing a new
value in the Variable or Observation field. The new tolerance limit will only apply to
the model generated by the current workset.
To change the tolerance limit of missing values for new worksets and models, see the
Project Options section in the File chapter.
Variables with zero variance
Variables are also checked for zero variance when clicking Check Missing Values or
OK. SIMCA issues a message for including or excluding the variable with zero
variance. When selecting to include such a variable it is given scaling weight 1.
Variables page
Use the Variables page, to view and configure the roles of the variables.
Available from the Variables page are:
1. Summary row listing Variables:
- Included - Number of variables included in the workset, number
of variables defined as X, number of variables defined as Y.
- Excluded - Number of excluded variables.
- Selected - Number of currently selected variables in the dialog.
2. Variable list displaying the current role of each variable (X, Y, or -), primary
variable ID, a Comment column, and a Dataset column when more than one
dataset was selected in the Select Data page.
3. Shortcut menu for selecting which variable IDs to display, Switch X & Y,
and commonly used commands. For more about Switch X & Y, see the
Switch X & Y subsection in the Change Model Type section later in this
chapter.
4. X for assigning variables to the X-block.
5. Y for assigning variables to the Y-block.
6. Exclude for excluding variables.
7. Class for assigning variables to the classes specified in the Observations
page.
8. Blocks for assigning variables to blocks for hierarchical modeling.
9. Phases - for assigning variables to phases (batch evolution model only).
10. Configure - for configuring the y-variable for the batch evolution model.
11. Find for finding variables. For more see the Find feature in Workset
page subsection previously in this chapter.
12. Save as Default Workset for saving the current specification as default
workset.
13. As model for assigning X and Y and excluding variables as another model.
See the As model subsection, previously in this chapter for more.
Home
139

Selecting roles by specifying X and Y
To change the variable roles:
1. Select the variables.
2. Click the desired button X, Y, or Exclude. The list is updated accordingly.
Note: Fitting a PLS, OPLS or O2PLS model with one qualitative Y is
equivalent to fitting a PLS-DA, OPLS-DA or O2PLS-DA model with classes
defined according to the qualitative variable.
Switch X & Y
Since O2PLS is bidirectional, it does not matter which variables are defined as X and
which are defined as Y. But, in the model window the statistics are displayed in one
direction.
To view the statistics for the other direction:
1. Click New as | Mxx and select the OPLS/O2PLS-model.
2. On the Variables page, right-click and select Switch X & Y. Note that all
variables previously defined as X are now defined as Y and vice versa.
3. Click OK and fit the new OPLS/O2PLS-model.

User Guide to SIMCA
140
Assigning variables to classes
When observations have been grouped in classes, by default all the X and Y variables
belong to all the classes.
To assign variables to specific classes:
1. Mark the variables.
2. Select the class in the Classes box.
The class number is displayed near the variable in the Class column.

Assigning variables to blocks resulting in hierarchical models
Variables can be assigned to variable blocks using the Blocks box.
To assign variables to blocks:
1. Mark the variables.
2. Select the block number from the Blocks box. After selecting the first block,
the Blocks list is incremented to include the next block number.
The block number is displayed near the variable under the column Class.
After defining the blocks the additional model types PCA-Hierarchical, and if y-
variables were specified, PLS-Hierarchical, OPLS-Hierarchical, and O2PLS-
Hierarchical are available. For more, see the Model type in the Workset dialog
subsection earlier in this section.
Home
141

Assigning variables to phases
When batches have phases, by default the X and Y variables belong to the phases as
specified at import.

To assign a variable to different phases than specified by the default workset:
1. Mark the variable.
2. Select the desired phases using the phase check boxes in the Phases box.
Note: Do not mark all variables that belong to a certain phase and select the
phase. Making another selection overlapping with the first will replace the
first phase setting for the intersection variables.
User Guide to SIMCA
142
Configuring y-variables
In a batch evolution model, the y-variable is a time or maturity variable. This time or
maturity variable can be configured differently depending on batch and y-variable
properties:
- Smooth - Smooth out noise in a maturity variable so that it is strictly
monotonically increasing or decreasing.
- Shift - Shifts batches and phases to always start at 0.
- Normalize - Batches are time normalized within each phase, giving all
batches in a model identical length.
To configure the y-variable, select it and click the Configure-button under Y Settings.
The following dialog opens:

Note: If the model should be used for on-line, never select the Normalize
check box in the Configure Y-Settings dialog.
Selecting variable type
When selecting to create a batch project, the y-variable type is Time or Maturity.
In the case where the project has batches but you want to just fit regular PLS models,
with Y not being time or maturity, select Normal (create a non-batch project) in the
Select variable type box.
Home
143
Selecting variable options
A combination or none of the three available options may be used for configuring y. In
cases where a y-variable relating to Batch age, starting at 0 at batch start, is available
no further y configuration is necessary. General recommendations for use of the three
options are described in the table.
Option Description Technical
details
Use Do not
use
Smooth Smoothes out
irregularities in the y
variable. This will
ensure that the
variable is strictly
monotonically
increasing or
decreasing, per batch
and phase.
Smoothing
by fitting a
constrained
quadratic
polynomial,
using a
piece wise
least
squares
algorithm.
When the y-
variable is a
maturity.
When the y-
variable is
time.
Shift Shifts all values so that
every batch and phase
starts at 0.
Shifting by
subtracting
the start
value of
each batch
from all
values.
Default
setting, few
exceptions.
When there
is missing
data at
beginning of
batches and
the y
variable is
Batch age.
Normalize Normalizes all batches
to the same length.
The y-values in every
batch and phase are
normalized into the
interval given by the
median start and
endpoints of the
batches.
Normalizing
by applying
a linear time
warping
function.
Rarely,
when batch
end is
chemically
or
biologically
driven.
For online
monitoring
models.
Save as Default Workset
To change the default workset roles of the variables, from what was specified at
import, specify the X, Y or excluded as desired and click Save as Default Workset.
This specification defines the default workset when clicking New in the Workset
group on the Home tab.
Save as Default Workset saves which datasets were selected. Deleting a dataset
included in the default workset results in a default workset using the other included
datasets.
For more about specifying the default workset, see the Default workset subsection
earlier in this chapter.
Observations page
The Observations page displays all observations for the current workset and is used
for including and excluding observations and grouping observations in classes.
User Guide to SIMCA
144
Available from the Observations page are:
1. Summary row listing:
- Observations: Total number of observations in the project, both
included and excluded.
- Included: Number of included observations.
- Selected: Number of currently selected observations in the dialog.
2. Observation list displaying for included and for excluded
observations, primary observation IDs, class number or name under Class,
and for batch projects Batch and Phase numbers under the respective
headers.
3. Shortcut menu for selecting which observation IDs to display and the
commands Include, Exclude, Set Class, and Select All.
4. Include for including excluded observations.
5. Exclude for excluding included observations.
6. Find for finding observations. For more see the Find feature in workset
dialog subsection previously in this chapter.
7. Find class for finding observations already assigned to a specific class.
8. Set class for assigning observations to a class number or entered name.
9. Class from obs ID for assigning classes from an available observation
identifier.
10. More for creating classes from variable or score values. For more, see the
Creating classes from variables or scores values subsection later in this
section.
11. As model for including, excluding, and assigning classes according to an
existing model. For more, see the As model subsection previously in this
chapter.
Note: The observations have to be included to be assigned to and part of a
class.

Home
145
Including and excluding observations
To include or exclude observations:
1. Select the observations.
2. Click the Include or Exclude-button.
For more about including and excluding observations, see the Excluding marked
items subsection in the Marked Items tab section in the Plot and list contextual tabs
chapter.
Grouping observations in classes using Set class
To group observations in classes using Set class:
1. Select the observations by marking or by using Find.
2. Specify the class by using either of the following methods:
- Clicking the Set button found after the Set class field. The class
number is automatically incremented.
- Typing the class name in the Set class field and clicking Set.
- Selecting an observation ID in the Class from obs ID and clicking
Set.

Automatic creation of classes from Observation ID
To automatically create classes from observations ID, primary or secondary:
1. Select the observation ID in the Class from obs ID box and click Set.

2. In the Set Class from Observation ID dialog, enter the starting position and
length of the string (number of characters) to be used for the creation of
classes.

User Guide to SIMCA
146
Creating classes from variables or scores values
To create classes from values of a variable or score vector from an existing model,
click the More button . The following window opens:

After selecting a variable or a score vector, the min, max, mean, and std. dev. fields
are automatically updated.
Find observations between section
To search for observations within a given range:
1. Click Variables or Scores.
2. Select the variable or score in the Variable or Score box.
3. Under Find observations between, type the limits defining the range for the
variable or score vector.
To assign the selected observations to a class, use Set class arrow and click the Set
button in the Observations page.
Auto-generate classes section
A set of classes can be generated automatically from variable or score values.
To generate classes automatically:
1. Click Variables or Scores.
2. Select the variable or score in the Variable or Score box.
3. Select number of classes from the Num. classes box.
4. Select Equal
a. Range Splitting the range of the selected variable or scores in the
desired number of classes. Note that this option may lead to
different size classes.
b. Size Splitting the group equally in the desired number of classes,
after sorting the variable. Note that this option may lead to
observations with identical values ending up in different classes.
5. Click the Set button in the More window.
Home
147
Auto-generating classes according to qualitative setting
When selecting a qualitative variable in the Variable box, clicking the Generate-
button automatically creates one class for each qualitative setting.

Observations page for batch evolution models
For batch evolution models the Observations page is stripped of all class setting
functionality. This means that observations belonging to one phase cannot be
reassigned to a different phase.
The Observation page can be used for viewing the observations, searching for
observations using the More-button or Find features, and to exclude or include single
observations.
To exclude entire phases or batches, or crop the phases, use the Batch page.
Batch page
The Batch page displays all batches and phases and is used for including, excluding,
and cropping batches and phases for batch evolution models, BEM. For batch level
models, BLM, the tab is there but no functionality is available.
Available from the Batch page are:
1. Summary row listing:
- Batches - total number of batches in the project.
- Included - Number of included batches in the current workset.
- Phases - Number of phases in the project.
2. Batch and phase lists displaying batch respective phase IDs and a Status
column. For batches, the Status column is updated with Exclude when
batches are excluded. For phases, the Status column states Included for
included phases and the number of observations currently included in the
phase. This column is updated when phases are cropped or excluded.
3. Shortcut menu with selected commands, also available as buttons.
4. Find for finding batches or phases. For more see the Find feature in
workset dialog subsection previously in this chapter.
User Guide to SIMCA
148
Under Batches
5. Include for including excluded batches.
6. Exclude for excluding batches.
Under Phases
7. Crop/Exclude for excluding observations phase-wise according to
specified criteria.
8. Uncrop/Include - for reversing the cropping.
9. Include All under Phases for including all phases.
10. Exclude for excluding phases.
11. Crop Information for viewing a summary of the performed cropping.

Note: When editing the workset (clicking Edit in the Workset group) and
selecting a phase (class) model, the workset dialog displays only the selected
model and the Batch page is not available.
Including and excluding batches
Below the batch list, the buttons Include and Exclude enable including and excluding
batches.
To exclude or include a batch:
1. Select the batch in the Batch list.
2. Click the Exclude or Include button. The Status column is updated and
states Excluded for the excluded batches and nothing for the included.
Note: Excluding batches excludes them from all phases.
Including and excluding entire phases
Below the phase list, the buttons Uncrop/Incl and Exclude enable including and
excluding entire phases.
Home
149
To exclude a phase:
1. Select the phase in the Phase list.
2. Click the Exclude button. The Status column is updated and states
Excluded for the excluded phases.
To include all phases, mark all phases and click the Uncrop/Incl button. The Status
column is updated and lists the number of observations in the respective phases.
Cropping phases
Crop a phase by excluding as specified in the Crop dialog.
Note: Batches with no phases are treated as having one single phase and can
be cropped in the Batch page.
To crop a phase:
1. Select the phase in the Phase list.
2. Click the Crop/Exclude-button.
3. Specify the desired cropping according to the table.
The different cropping options are described in the table.
Option Description Example and
result
1. Exclude the first x
observations in all
batches in this
phase.
Excludes x
observations, starting
from the beginning of
each batch, in the
marked phase.
Entering 3 excludes the
3 first observations in all
batches for the marked
phase.
2. Exclude the last x
observations in all
batches in this
phase.
Excludes x
observations, starting
from the end of each
batch, in the marked
phase.
Entering 3 excludes the
3 last observations in all
batches for the marked
phase.
3. Exclude all
observations
satisfying the
following condition:
If variable VAR1 is
LOGICAL
EXPRESSION
VALUE1 and
VALUE2.
Excludes the
observations satisfying
the condition defined
by:
- The variable.
- The logical
expression
among <, >,
<=, >=, and
outside.
- The limiting
value or for
outside values.
If variable bottom temp.,
is <, 60 excludes all
observations with a value
smaller than 60 in
variable bottom temp.
Note that the result from
cropping all phases with
this type of cropping can
be achieved with
trimming too.
4. Downsize, use only
every x observation.
Downsizes the phase
to only use every x
observations.
Entering 3 excludes 2/3
of the observations in the
phase, keeping
observation number 1, 4,
7, 10, 13 etc.
User Guide to SIMCA
150
Note: The check boxes can be selected independently and all together.

Crop Information
After cropping, clicking the Crop Information button displays a summary of the
performed cropping per phase and per cropping. This means that for each time OK is
clicked with cropping defined, the applied cropping details are displayed per phase in
the Crop Information window.
Transforming variables
The Transform page displays all variables with current transformation and statistics.
The Transform page is used to apply transformations to selected variables.
Available from the Transform page are:
1. Summary row listing number of included variables and number of selected
variables.
2. Variable list displaying variable role (X or Y), primary variable ID,
Skewness, Min/Max, Min, Max, Transform, and the constants C1, C2, and
C3. For more about the Transform, C1, C2, and C3 columns see the
Applying transformations subsection later in this section.
3. Shortcut menu with the commands for selecting which variable labels to
display, Set Transform, Select All, and Quick Info.
4. Find for finding variables. For more see the Find feature in workset
dialog subsection previously in this chapter.
5. Quick Info for opening the Quick Info window. For more see the Quick
Info subsection in the View chapter.
6. As model for transforming all variables as transformed in another model.
See the As model subsection, previously in this chapter for more.
7. Specify transformation for defining a transformation and applying it to
the selected variables.
8. Auto transform selected variables as appropriate to automatically
transform the variables in need of logarithmic transformation.
Note: When batch data have phases, transformations of variables can be done
for each phase by editing the phase model (edit by marking the phase model
and on the Home tab, in the Workset group, clicking Edit).
Home
151

Skewness or Min/Max colored red
When the value in Min/Max or Skewness is colored in red, a transformation is
probably desired.

For more about the criteria used, see the Transform page criteria section in
Appendix A: Statistics.
Quick Info in Transform page
With the Quick info window open when applying transformations, the statistics and
plots displayed in the Quick Info are updated. This enables you to view the before and
after plots and statistics immediately.
To open the Quick Info window, click the Quick Info button.
For more about the Quick Info window, see the Quick Info subsection in the View
chapter.
User Guide to SIMCA
152
Applying transformations using Specify transformation
To apply a transformation:
1. Select the variables to transform in the list.
2. Select a transformation in the Specify transformation box. Under the
selected transformation, the formula of the selected transformation appears
next to Formula. The default values for the constants appear in the C1, C2,
and C3 fields.
3. Change from the default constants as appropriate.
4. Click the Set-button. The columns Transform, C1, C2, and C3 in the
variable list are updated accordingly.
5. Repeat 1- 4 as needed to transform all the desired variables.
Note: The transformation is applied when clicking the Set button. Clicking
OK without having clicked the Set button does not apply the selected
transformation.
Transformations available
The following transformations are available:

Automatic transformation
Automatic Log transformation according to certain criteria is available by following
the steps:
1. Select the desired variables in the variable list.
2. Optionally select the If one variable in the X or Y block needs
transformation, transform all selected variables in that block check box
to have all variables in the same metric. In summary, when the check box is
selected all variables are Log transformed if one meets the criteria for
automatic transformation.
3. Click the Transform button. The software checks if any of the selected
variables need a Log transformation and if so applies it.
The criteria used to decide if a log transform is needed is described in the Transform
page criteria section in Appendix A: Statistics.
Home
153
Lagging variables
When the X or Y-matrix is expanded with lagged variables, this allows the study of the
influence of the process variables at L time units earlier in the process at time t.
All variables, included or excluded, can be lagged and included in the model.
Available from the Lag page are:
1. Lags field for defining lagging structure.
2. The lists displaying variable role and primary variable ID for Available and
Lagged variables.
3. Shortcut menu with commonly used commands.
4. Arrow button between the fields to add lagged variables.
5. Find for finding variables. For more, see the Find feature in workset
dialog subsection previously in this chapter.
6. Remove and Remove All for removing lag variables.
7. As model for adding lagged variables as in another model. For more, see
the As model subsection previously in this chapter.

Lag specification
A variable can be lagged with several different lags, e.g., 2, 4, 12, 24 and 36.
In Lag specification, enter the desired lags by typing. Use to specify a range, for
example 1-5 means lags 1 to 5. After defining a lag structure, that structure can be
selected by clicking the Lags arrow.
Adding the lagged variables
To create the lagged variables:
1. Enter the lag structure in the Lags field.
2. Mark the variables in the Available variables list.
3. Click the arrow . The lag variables are added in the Lagged variables
list, according to the selected lag structure.
To remove any of the new lagged variables, mark it and click Remove.
User Guide to SIMCA
154
Expanding variables
The Expand page allows changing the default linear model by expanding the X matrix
with squares, cross, and cubic terms.
Available from the Expand page are:
1. Summary row listing number of included variables, selected variables,
expanded terms, and selected expansions.
2. Variables list displaying the included variables, including lag variables, and
their roles.
3. Expanded terms list displaying the created expansions.
4. New term field and => buttons for defining expansions. For more see the
Adding expansions using the New term field subsection later in this
chapter.
5. Shortcut menu for selecting which variable IDs to display and commonly
used commands.
6. Cross, Square, Sq & Cross, and Cubic-buttons between the fields to add
expansions. For more see the Adding terms using the buttons subsection
later in this chapter.
7. Find for finding variables. For more see the Find feature in workset
dialog subsection previously in this chapter.
8. Remove and Remove All for removing expansions.
9. As model for adding expansions as in another model. See the As model
subsection, previously in this chapter for more.

Term names
The names for expanded term are 'primary variable ID' x 'primary variable ID', i.e.
x1*x1 for squares, x1*x2 for cross terms, and x1*x1*x1 for cubic terms. These names
will appear on all plots and lists.
Scaling of expanded terms
All expanded terms are displayed in the Scale page. For more, see the Scaling of
expanded terms subsection in Appendix A: Statistics.
Home
155
Adding terms using the buttons
The x-variables, including lagged variables, are displayed in the Variables list.
To add an expansion, select the desired variables and click a button:
Adding terms using the New term field
Any term, including up to three of the available variables, can be added using New
term.
To add three factor interactions:
1. Add a term (and see that it ends up in New term) by:
a. Double-clicking a term OR
b. Selecting a term and click the left arrow =>.
2. Repeat 1. until the three factor interaction is displayed in New term.
3. Click the rightmost => arrow to add the term.
To empty the New term field, click the Clear button.
Note: The New term field is useful when the term you want to add includes
three terms but is not cubic.
Scaling variables
Usually with all variables of the same type or an assortment of different measurements,
centering and auto scaling to unit variance is warranted. This is the SIMCA default
and if this is your choice, no changes in the Scale page are necessary. The SIMCA
default can be changed in Project Options, see the Fitting options subsection in the
Project Options section in the File chapter for more.
Customizing the scaling is available from the Scale page in 4 manners:
1. Using the Set Scaling section.
2. Importing scaling weight and center values from file by clicking the Read
Scaling button.
3. Defining the scaling weight and center values from secondary variable ID by
clicking the Scale from Sec. ID button.
4. Defining the scaling weight and center values by entering them manually in
the dialog by clicking the Custom Scaling button.
When using customized scaling (2-4 above), variables with values left blank for the
scaling weight use the calculated standard deviation from the data to calculate the
weight. Similarly, variables with values left blank for the center use the calculated
mean of the variable as center.
Note: To display the current scaling weight and center values after using
Read Scaling or Scale from Sec. ID, click the Custom Scaling button.
User Guide to SIMCA
156

Centering or not centering
Normally, variables are centered by subtracting their averages or other reference
values. Centering by subtracting the average is the SIMCA default.
Occasionally, for example when working with difference data, you may be interested
in the variation of the variables around zero (0), hence you do not want to center
specific variables. Use one of the customization methods listed above to accomplish
this.
Note: The scaling weights of all variables, except those with base type Freeze
or Frozen, are recomputed when changing the selection of observations.
Show statistics
Selecting or clearing the Show statistics check box displays or hides the average and
standard deviation of every variable.
Note: When using Read Scaling, Scale from Sec. ID, or Custom Scaling the
values displayed under Avg and Std. Dev. are the calculated average and
standard deviation values, not the values that will be used when scaling the
variables.
To read about the scaling of expanded terms, transformed variables, lagged variables,
variables in classes, scaling after re-selecting observations, and calculation of the
scaling weight, see the Scaling section in Appendix A: Statistics.
Scaling using the Set Scaling section
To change the scaling base type from the Set Scaling section:
1. Select the variables.
2. Select the base type in the Type box.
3. Click Set next to the Type box.
The new scaling type is displayed in the Type column.
Home
157

Scaling base weight types available
The following scaling base types are available from the Type box:
Base
weight
type
Description
None No centering or scaling (ws = 1).

UV Variable j is centered and scaled to "Unit Variance", i.e. the base
weight is computed as 1/sd
j
, where sd
j
is the standard deviation of
variable j computed around the mean.
UVN Same as UV, but the variable is not centered.
The standard deviation is computed around 0.
Par Variable j is centered and scaled to Pareto Variance, i.e. the base
weight is computed as 1/sqrt(sd
j
), where sd
j
is the standard deviation of
variable j computed around the mean.
Pareto scaling is in between no scaling and UV scaling and gives the
variable a variance equal to its standard deviation instead of unit
variance.
ParN Same as Par, but the variable is not centered
The standard deviation is computed around 0.
Ctr The variable is centered but not scaled (ws = 1)
Freeze The scaling weight of the variable is frozen and will not be re-computed
when observations in the workset change or the variable metric is
modified after the freezing.
Block scaling
Block-wise scaling is warranted when a data table contains several types (blocks) of
variables, with different numbers of variables in each block. Block-wise scaling allows
each block to be thought of as a unit and to be given the appropriate variance, which is
less than if each variable was auto scaled to unit variance.
Block scaling scales down the importance of the variables so that the whole block has
\(Kblock) variance, or unit variance,
where Kblock is the number of variables in the block.
User Guide to SIMCA
158
Defining the block scaling
To block scale:
1. Mark the variables in the first block.
2. Select the block weight.
3. Click the Set button found to the right of the blocking options.
4. Mark other variables.
5. Change the block number and optionally the block weight.
6. Click the Set button found to the right of the blocking options.
7. Repeat 4-6 until all blocks have been defined.
Block scaling types available
The following block scaling types are available:
where Block = number of variables in the block
Note: X and Y variables cannot be grouped in the same scaling block.
Modifier
Scaling variables up or down relative to their base weight is done using the Modifier.
To scale variables up or down relative to their base weight:
1. Mark the variables.
2. Enter the value in Modifier.
3. Click Set.
To reset the modifier to 1:
1. Mark the variables.
2. Enter 1 in Modifier.
3. Click Set.
Reading scaling from file
Centers and scaling weights can be imported from file and used to scale the variables.
Importing scaling from file for regular projects
For regular projects, i.e. non batch data, the file must have the following rows:
- Variable IDs.
- Center values.
- Scaling weights.
Home
159
Importing scaling from file for batch projects with phases
Specifying different scaling than the default for the variables in the different phases
and applying them simultaneously to all phases, is available using Read Scaling.
For batch projects with phases, centers and scaling weights for each phase can be
defined in the Read Scaling wizard spreadsheet.
The imported file must have the following column and rows:
- Phase ID column.
- Variable names row.
- Center values rows, one for each phase.
- Scaling weights rows, one for each phase.
To see the different base weight types resulting from customized scaling, see the
Scaling base weight types with customized scaling section later in this chapter.
Note: When using Read Scaling to specify the scaling for all phases
simultaneously, either click Edit | BEMxx or New in the Workset group.
Using Read Scaling when editing a phase will apply the scaling to that phase
only. This is the case for class models too.
Read Scaling wizard
To import a file holding scaling weights and center values follow the steps that follow:
Step Action and illustration
1. Select the scaling file. Click Read Scaling and select the file containing the
weights and/or centers.

User Guide to SIMCA
160
Step Action and illustration
2. Specify weights and
centers.
For batch projects with
phases, weights and
centers need to be specified
for all phases.
In the Import Scaling Data dialog, in turn mark the
rows with primary variable IDs, weights, and centers
and click the Variable IDs, Weights and Center-
buttons.
SIMCA automatically recognizes the words mean or
center and weight when present in a column and
marks these rows as center and weights.
Variables that are not included in the list are centered
and scaled UV. Variables with blank values for the
center use as center the computed mean from the
data and variables with blank for scaling weights are
scaled UV.
Note: When the batch project has phases, weights
and centers have to be specified for each phase.
Click Next.


Home
161
Step Action and illustration
3. Specifying weight
properties.
In the Summary page, specify how to use the
weights and click Finish.
View the table next for a description of the options in
the Summary page.


User Guide to SIMCA
162
Weight definition for imported scaling
This table describes the options in the Import Scaling Data wizard Summary page.
Weight selected Description Example
The standard deviation
of the variable. SIMCA
will use the inverse of
the values as scaling
weights.
The values specified as
Weights in the
spreadsheet are inverted
and used as scaling
weight.
If the Weights value for
variable V1 in the
spreadsheet is 3, then
the scaling weight, ws =
1/3.
The inverse of the
standard deviation of
the variable. SIMCA
will use the values as
the scaling weights.
The values specified as
Weights in the
spreadsheet are used as
scaling weight.
If the Weights value for
variable V1 in the
spreadsheet is 3, then
the scaling weight, ws =
3.
% of the center values.
SIMCA will use the
inverse of the % center
values as scaling
weights.
The values specified as
Weights in the
spreadsheet are the
percentage values that,
multiplied with the center
values, results in the
standard deviation, SD (the
inverse of the scaling
weight).
When Centers are
specified in the
spreadsheet, the SD used
is the percent (weight
value) of the center.
For variables with the
Centers left blank, the
calculated mean is used as
center and the standard
deviation (SD) used is the
specified percentage of the
center or the calculated
SD, whichever is greater.
If the Weights are blank,
this option defaults to using
the SIMCA calculated
weight.
With center value: If the
Weights value for
variable V1 in the
spreadsheet is 3 and
the Centers value is 10,
then the scaling weight,
ws = 1/(0.03*10).
Without center value: If
the Weights value for
variable V1 in the
spreadsheet is 3 and
the Centers value is
blank, then the scaling
weight used is the
smaller of
ws = 1/(0.03*calculated
center)
ws = 1/(the calculated
SD).
Use xx % of the center
values as standard
deviation when
calculating the weight.
The values defined as
Weights in the
spreadsheet are ignored.
The entered percentage of
the center values is used
as standard deviation and
inverted before used as
standard deviation.
If xx entered is 5, the
center value is 10. The
standard deviation, SD =
0.05*10. The scaling
weight, ws = 1/SD = 2.
Note: When using Read Scaling to specify the scaling for all phases
simultaneously, either click Edit | BEMxx or New in the Workset group.
Using Read Scaling when editing a phase will apply the scaling to that phase
only.
Home
163
Scaling from numerical secondary variable ID
Scaling from secondary variable IDs is available when secondary variable IDs are
numerical. The variable IDs must contain scaling weights or center values.
Note: When batch data have phases, scaling from Secondary IDs can be done
for each selected phases, by editing the phase model.
To see the different base weight types resulting from customized scaling, see the
Scaling base weight types with customized scaling subsection later in this chapter.
Scale from Sec. ID dialog
To scale and/or center from secondary variable IDs, follow the steps:
1. Select the variables to scale.
2. Click the Scale from Sec. ID button.
3. In the dialog that opens, select the secondary variable ID that contains the
standard deviations or scaling weights in the Weight box. Selecting 1 (no
scaling) results in no scaling.
4. In The weight is section, the default is The standard deviation of the
selected variables. This means that the inverted values in the variable ID
will be used as scaling weights. To use the specified weight values in the
secondary ID as scaling weight, select The inverse of the standard
deviation for the selected variables.
5. In the Center box, select the secondary ID that holds the center values or 0
(center around zero).
6. Selecting the Use x % of the center as standard deviation when
calculating the weight check box results in that the entered percentage of
the center values is used as standard deviation and inverted before used as
scaling weight. When selecting this check box the other Weight options are
unavailable. For more details, see the second table in the Read Scaling
wizard subsection earlier in this chapter.

User Guide to SIMCA
164
Custom Scaling dialog
It is sometimes necessary to specify the center and the scaling weight of variables to be
different than computed from the data. By clicking the Custom Scaling button, the
scaling weight of the variables can be specified as follows:
1. Select the variables to scale.
2. Click the Custom Scaling button.
3. In the dialog that opens, enter the standard deviation or scaling weight in the
Weight field. Leaving blank results in using the calculated scaling weight.
4. In The entered weight is section, the default is The standard deviation of
the selected variables option. This means that the inverted values in the
variable ID will be used as scaling weights. To use the specified weight
values in the secondary ID as scaling weight, select The inverse of the
standard deviation for the selected variables option.
5. In the Center field, enter the center value to use. Leaving blank results in
using the calculated mean as center.
6. Selecting the Use x % of the center as standard deviation when
calculating the weight check box results in that the entered percentage of
the center values is used as standard deviation and inverted before used as
scaling weight. When selecting this check box the other Weight options are
unavailable. For more details, see the second table in the Read Scaling
wizard subsection earlier in this chapter.

Scaling base weight types with customized scaling
After changing the scaling using Read Scaling, Scale from Sec. ID, or Custom
Scaling the following scaling base types are available:
Base
weight
type
Description Weight and center
specification
Frozen The center and the scaling weight
of the variable are specified
Weight and Center values that
are not weight=1 and center=0.
Home
165
Base
weight
type
Description Weight and center
specification
FrozenC The scaling weight is specified,
but not the center (value is left
blank). The center is computed
from the data.
Weight values (not=1) but
Center is left empty.
Not possible for Scale from
Sec ID.
FrozenN The scaling weight is specified
and no center is selected,
(centering is done around 0).
Weight values (not=1) and
Center = 0.
UVF The center is specified, but not
the scaling weight, the standard
deviation is computed from the
data and ws=1/standard
deviation. UVF is UV scaled using
a new center value.
Weight values left empty,
Center values specified.
Not possible for Scale from
Sec ID.
NoneF The center is specified and no
scaling is selected (ws=1)
Weight = 1, Center values
specified.
%Mean The center is not specified but the
entered percentage of the center
is used as standard deviation
when calculating the weight.
No center or weight values
specified. The Use x % of the
center as standard deviation
when calculating the weight
check box selected and the
desired percentage entered.
%Frozen The center is specified and the
entered percentage of the center
is used as standard deviation
when calculating the weight.
Center specified but not weight
values. The Use x % of the
center as standard deviation
when calculating the weight
check box selected and the
desired percentage entered.
The calculation of the scaling weight is described in Scaling weight calculation
subsection in Appendix A: Statistics.
Spreadsheet
The Spreadsheet page, in the Workset dialog, displays the currently included
variables and observations, including lagged and expanded variables.
Available in the Spreadsheet page are:
- Quick Info-button for viewing variables and observations.
- Reset limits-button for removing trimming and Winsorizing.
- View box for selecting scaled or original units. When editing a batch
model, average batch and aligned vectors are available.
- Trim Information-button for displaying the performed trimming.
- Workset list displaying all variable and observation IDs and values.
- Shortcut menu for copying, locking columns, opening the Quick Info,
zooming out, saving list to file, and printing.
User Guide to SIMCA
166

The Spreadsheet list displays the data and all variable and observation IDs of the
currently included variables and observations.
Quick Info in the Workset dialog spreadsheet
The Quick Info window displays overview information about the marked variables or
observations, depending on whether selecting Quick Info | Variables or Quick Info |
Observations.
Open the Quick Info window by:
- Clicking the Quick Info button and selecting Variables or Observations.
- Right-clicking the spreadsheet and selecting Quick Info | Variables or
Observations.
Trim-Winsorizing the workset
When trimming or Winsorizing the workset, it only affects the current workset and
does not touch the dataset.
The Trim-Winz Var and Trim-Winz All buttons are available in the Quick Info
Variables window when workset as raw data is selected in the View box, that is,
when the variables are displayed in original units.
Trimming and Winsorizing is done on the selected variables across all phases and
batches.
Note: When editing a phase model, trimming is unavailable.
For more about the Quick Info window, see the Quick Info section in the View
chapter.
Removing the trimming and Winsorizing in the workset
Removing the trimming or Winsorizing for the current workset can be done by:
- Clicking the Undo trimming button in one of the trimming Winsorizing
dialogs.
- Clicking the Reset limits button next to the Quick Info button.
Home
167
Trimming and Winsorizing information
After trimming or Winsorizing, clicking Trim Information displays a summary of the
performed trimming/Winsorizing.

View workset in scaled or original units
The View box contains the two options workset and workset as raw data and when
editing a phase or batch model the additional options: average batch, and aligned
time/maturity vector.

By default, workset is selected in the View box, meaning that the variables are
transformed, scaled, and centered according to the current workset specification.
To display all variables and observations in original units, select workset as raw data.
Average batch and aligned time/maturity
To view the average batch or aligned time or maturity vector for a BEM with phases,
mark a phase model and click Edit in the Workset group. In the Spreadsheet page of
the Workset dialog select average batch or aligned time/maturity vector in the
View box. When editing a BEM (with phases) or BLM the additional batch options
are unavailable.
Note: With phases, these calculations are done for every phase, since the
alignment is done by phase.
Creating unfitted models
New unfitted models are created when exiting the Workset dialog by clicking OK.
Unfitted models in wrappers
When the Model type is PCA/PLS/OPLS/O2PLS-Class SIMCA generates unfitted
class models, one for each class with the wrapper CMxx. The numbering of the CM
models is sequential.
For batch projects with phases, when clicking OK SIMCA generates unfitted PLS
class models, one for each phase with the wrapper BEMxx.
User Guide to SIMCA
168
Note: Changes in the class or phase model are not remembered by the
wrapper model CM respective BEM. This means that if you for instance
exclude a batch in a phase model, that batch will be included if you click
Workset | New As Model and click the BEM.
Simple mode workset wizard
The workset dialog is opened in simple mode the very first time you open the Workset
dialog in SIMCA. After that the dialog remembers whether to open in simple mode or
advanced mode.
The simple mode workset wizard contains the following pages:
1. Start page displays a short description of the workset wizard.
2. Select Data page - displays the available datasets when there is more than
one. See also the Select Data subsection earlier in this chapter.
3. Variables page for assigning roles and applying automatic transformation.
4. Observations page for excluding, including, and assigning classes.
5. Summary page displays a summary of the active model and allows
changing from the default fitting type.
To switch between simple and advanced mode, click the Use Advanced Mode and
Use Simple Mode respectively.
Note: Simple mode is only available for regular projects.
Simple mode start page
The first page of the workset wizard describes the features of the workset wizard.
To not display this introductory page the next time the simple mode workset opens for
this project, select the Dont show this page again check box.
Click Next.

Home
169
Variables page in simple mode
In the Variables page, the following is available:
1. A summary line displaying currently included and excluded variables.
2. The variable list displaying the current role of each variable (X, Y, or ' -'),
variable IDs, and a Comment column.
3. X-button for assigning variables to the X-block.
4. Y-button for assigning variables to the Y-block.
5. Exclude-button for excluding variables.
6. Find and select field for finding variables. For more see the Find feature
in Workset page previously in this chapter.
7. Save as Default Workset button for saving the current specification as
default workset.
8. As model for assigning X and Y and excluding variables as an existing
model. See the As model section, previously in this chapter for more.
Clicking Next opens the Observations page of the wizard.

Variable list
The variable list displays the current role, variable IDs, and Comment. The Comment
column displays the current transformation. The only transformation available in
Simple Mode is the Log transform.
The following can be added in Advanced Mode and is then displayed in Simple
Mode:
- Transformation other than Log displayed in Comment.
- Lagging displayed in Comment.
- Class belonging displayed in Class.
The context sensitive menu, opened by right-clicking the variables list, holds the
following commands: X, Y, Exclude, Select All (CTRL+A), and Variable Label,
where the last enables displaying the desired variable IDs.
User Guide to SIMCA
170
Selecting roles by specifying X and Y
To change the variable roles:
1. Select the variables.
2. Click the desired button X, Y, or Exclude. The list is updated accordingly.
Note: Fitting a PLS, OPLS or O2PLS model with one qualitative Y is
equivalent to fitting a PLS-DA, OPLS-DA or O2PLS-DA model with classes
defined according to the qualitative variable.
Automatic transformation
Automatic Log transformation according to certain criteria is available by following
the steps:
1. Select the desired variables in the variable list.
2. Optionally select the If one variable in the X or Y block needs
transformation, transform all selected variables in that block check box
to have all variables in the same metric. In summary, when the check box is
selected all variables are Log transformed if one meets the criteria for
automatic transformation.
3. Click the Transform button. The software checks if any of the selected
variables need a Log transformation and if so applies it.
The criteria used to decide if a log transform is needed is described in the Transform
page criteria section in Appendix A: Statistics.
Save as Default Workset
To change the default workset roles of the variables, from what was specified at
import, specify the X, Y or excluded as desired and click Save as Default Workset.
This specification defines the default workset when clicking New in the Workset
group on the Home tab.
Save as Default Workset saves which datasets were selected. Deleting a dataset
included in the default workset results in a default workset using the other included
datasets.
For more about specifying the default workset, see the Default workset subsection
earlier in this chapter.
Home
171
Observations page in simple mode
The Observations page functionality is identical to the Observations page in the
Advanced Mode with the following exception: there is no More button so assigning
classes from variables or scores is unavailable here.
For details about the Observations page, see the Observations page section
previously in this chapter.
Click Next to open the Summary page.

Summary page in simple mode
The Summary page displays a summary over the specification done in the workset
wizard, including the model type that will be fitted.

Click Finish to close the wizard and automatically fit the model or models.
Change Model Type
The default model type after exiting the workset dialog is determined by the selected
model type in the Model Type box in the Workset dialog.
User Guide to SIMCA
172
To change from the default model type, click Change Model Type and click a new
model type in the gallery that opens.
For example after fitting a principal component model of the responses (PCA-Y),
click Change Model Type and select PLS to fit a PLS model. SIMCA creates a
new unfitted model, with type PLS.
An unfitted model can only be a Non Hierarchical Base Model.
Note: SIMCA creates a new unfitted model every time a new model type is
selected.

The Change Model Type button is found in the Workset group on the Home tab.
Model types available
The available model types from Change Model Type are, in the order of the gallery:
Model type Description
Overview
PCA-X Fits a PC to the X variables.
PCA-Y Fits a PC to the Y (responses) variables.
PCA-X&Y Fits a PC to all included variables (X and Y).
Regression
PLS Fits a PLS model.
Available when both X and Y variables were defined in the workset.
Home
173
Model type Description
OPLS Fits an OPLS model with the additional objective to model and
interpret variation in X that is orthogonal to Y, which needs to be
modeled to achieve the best possible prediction and interpretation
Available when both X and Y variables were defined in the workset.
O2PLS Fits an O2PLS model with the additional objective to model and
interpret variation in X that is orthogonal to Y.
The orthogonal variation can be further divided into two parts; the
OPLS part and the PCA part. The PCA part consists of structured
variation that does not affect the prediction but can be interesting to
study to further improve the interpretation of complex multivariate
data.
Available when X and at least two Y variables were defined in the
workset.
Discriminant Analysis
PLS-DA Fits a PLS model using SIMCA created dummy Y variables, one for
each class.
Available after grouping observations in two or more classes.
OPLS-DA Fits an OPLS/O2PLS model using SIMCA created dummy Y
variables, one for each class. See OPLS above.
Available after grouping observations in two or more classes.
O2PLS-DA Fits an O2PLS model using SIMCA created dummy Y variables,
one for each class. See O2PLS above.
Available after grouping observations in three or more classes.
Class Models
PCA-Class Fits PC models for the X block, one for each class, with a wrapper
CM.
Available after grouping observations in classes or variables in
blocks.
PLS-Class Fits PLS models, one for each class, with a wrapper CM.
Available after grouping observations in classes, or variables in
blocks, and both X and Y variables are included.
OPLS-Class Fits OPLS models, one for each class, with a wrapper CM. See
OPLS above.
Available after grouping observations in classes, or variables in
blocks, and both X and Y variables are included.
O2PLS-Class Fits O2PLS models, one for each class, with a wrapper CM. See
O2PLS above.
Available after grouping observations in classes, or variables in
blocks, and both X and at least two Y variables are included.
Clustering
PLS -Tree Fits several PLS models grouped as branches in a tree.

User Guide to SIMCA
174
Note: DA models are fitted to the X variables only. The Y-variables
(responses), when present, are ignored.

Note: All results of the fit of a PC model (PCA-X and PCA-Y), such as SS
explained are labeled as X (i.e. R2X).

Note: A PLS, OPLS or O2PLS model with one qualitative y-variable is
equivalent to PLS-DA, OPLS-DA and O2PLS-DA respectively.
Model window for OPLS and O2PLS models
The Model Window corresponding to OPLS and O2PLS displays summary statistics
related to the model. Components that capture variation found in both X and Y are
denoted Predictive. Components that capture variation only found in X are denoted
Orthogonal in X(OPLS). Components that capture variation only found in Y are
denoted Orthogonal in Y(OPLS).
The OPLS model is based on the OPLS algorithm. The O2PLS model is based on the
similar O2PLS algorithm, but in addition there is a PCA step. PCA is used after
convergence of the O2PLS algorithm, to exhaust the E and F residual matrices from all
remaining systematic variation. This yields the additional Orthogonal in X(PCA) and
Orthogonal in Y(PCA) components.
The method underlying a certain Orthogonal component is indicated in the name of the
component in the Model Window.
Note: The model window for OPLS models displays, if existing, three types of
components: Predictive, Orthogonal in X, Orthogonal in Y. Additionally, The
model window for O2PLS displays, if existing, Orthogonal in X and
Orthogonal in Y components estimated by PCA.
Home
175
Figure 2. The Model Window for an O2PLS model with 7 predictive, 3 Orthogonal in
X and 6 Orthogonal in Y components (a 7+3(1+2) +6(1+5) O2PLS model).

The Model Window displays:
Section Description Component types
Model Summarizes the model,
showing the cumulative
R2X, R2, Q2, and R2Y.

Predictive The Predictive section
where the first row
summarizes the predictive
components in the model
followed by a listing of
each predictive
component.
The predictive loading vectors are p
for the X-block and q for the Y- block.
The predictive score vectors are t for
the X-block and u for the Y-block. In
the figure above there are 7
components for p, q, t and u.
Orthogonal
in X (OPLS)
The Orthogonal in
X(OPLS) section where
the first row summarizes
the orthogonal
components in the X
model followed by a listing
of each orthogonal in X
component. The
orthogonal in X
components show the
variation in X that is
uncorrelated to Y.
The orthogonal in X(OPLS) loading
vectors are po for the X-block and so
for the Y-block. The orthogonal in X
(OPLS) score vector is to for the X-
block.
In the figure above there is one
orthogonal in X(OPLS) component.
This means that the po[1], so[1], and
to[1] vectors are the orthogonal in
X(OPLS) vectors.
User Guide to SIMCA
176
Section Description Component types
Orthogonal
in X (PCA)
The Orthogonal in X(PCA)
sections where the first
row summarizes the
orthogonal components in
the X model followed by a
listing of each orthogonal
in X component. The
orthogonal in X
components show the
variation in X that is
uncorrelated to Y
The orthogonal in X(PCA) loading
vectors are po for the X-block and so
for the Y-block. The orthogonal in X
(PCA) score vector is to.
The orthogonal in X(PCA)
components are extracted after the
orthogonal in X(OPLS) components,
and their relational order is indicated
through the shared nomenclature.
In the figure above there are two
orthogonal in X(PCA) components.
This means that the po[2], po[3],
so[2], so[3], to[2] and to[3] vectors
are the orthogonal in X(PCA) vectors
Orthogonal
in Y (OPLS)
The Orthogonal in Y
(OPLS) sections where the
first row summarizes the
orthogonal components in
the Y model followed by a
listing of each orthogonal
in Y component. The
orthogonal in Y
components show the
variation in Y that is
uncorrelated to X.
The orthogonal in Y(OPLS) loading
vectors are r for the X-block and qo
for the Y-block. The orthogonal in Y
(OPLS) score vector is uo for the Y
block.
In the figure above there is one
orthogonal in Y(OPLS) component.
This means that the r[1], qo[1], and
uo[1] vectors are the orthogonal in
Y(OPLS) vectors.
Orthogonal
in Y (PCA)
The Orthogonal in Y (PCA)
sections where the first
row summarizes the
orthogonal components in
the Y model followed by a
listing of each orthogonal
in Y component. The
orthogonal in Y
components show the
variation in Y that is
uncorrelated to X
The orthogonal in Y(PCA) loading
vectors are r for the X-block and qo
for the Y-block. The orthogonal in Y
(PCA) score vector is uo for the Y
block.
The orthogonal in Y(PCA)
components are extracted after the
orthogonal in Y(OPLS) components
and their relational order is indicated
through the shared nomenclature.
In the figure above there are five
orthogonal in Y(PCA) components.
This means that the r[2]-r[6], qo[2]-
qo[6], and uo[2]-uo[6] vectors are the
orthogonal in Y(PCA) vectors.
The table columns are:
- Component - Component index.
- R2X - Fraction of X variation modeled in that component, using the X
model.
- R2X(cum) - Cumulative R2X up to the specified component.
- Eigenvalue - The minimum number of observations (N) and X-variables
multiplied by R2X, that is, min(N,K)*R2X.
- R2 - Fraction of Y variation modeled in that component, using the X model.
- R2(cum) - Cumulative R2 up to the specified component.
Home
177
- Q2 - Fraction of Y variation predicted by the X model in that component,
according to cross-validation.
- Q2(cum) - Cumulative Q2 up to the specified component.
- R2Y - Fraction of the Y variation modeled in that component, using the Y
model.
- R2Y(cum) - Cumulative R2Y up to the specified component
- EigenvalueY - The minimum number of observations (N) and Y-variables
multiplied by R2Y, that is, min(N,M)*R2Y.
- Significance Significance level of the model component.
For more, see the OPLS/O2PLS - Orthogonal PLS modeling section in Appendix
A: Statistics.
Model Options
The Model Options hold model specific options. Most options are inherited from the
Project Options.
Note: Model options cannot be changed for a wrapper model, only for the
individual class or phase models.
To open the Model Options:
- Click the Model Options dialog dispatch launcher in the bottom right corner
of the Workset group, on the Home tab.
- Click the Options button in the Model Window.
- Right-click the model, either in Project Window or in Model Window, and
click Model Options.
Model Options contain the following pages:
- Model for setting confidence and significance levels, and selecting whether
to use cross validation when fitting.
- Distance to Model for selecting in which domain the distance to model
should be displayed.
- Coefficients for selecting type of coefficient to display by default.
- Residuals / R2 for selecting residual and R2 type to display by default.
- Predictions for selecting to display the predictions in scaled and
transformed units and whether to trim the predictions as the workset.
- CV-groups enables customized assignment of the cross validation groups.
- More Options displays and allows change of the used T2 Center and
Eigenvalue similarity level.
Note: Changes in Model Options are only valid for the current model.
User Guide to SIMCA
178
Model page in Model Options
The Model page displays:
- Use cross validation when fitting box.
- Confidence level on parameters current setting.
- Significance level for DModX and Hotellings T2 current setting.
- Reset All button.

Cross validation when fitting
By default, cross validation is used when fitting to decide whether a component is
considered significant or not.
To fit components without cross validation, clear the Use cross validation when
fitting check box.
Autofit is unavailable when the Use cross validation when fitting check box is
cleared.
Confidence level on parameters
The confidence level used when computing confidence intervals on the parameters is
by default 95%.
To change from the default select 90%, 99%, or None in the Confidence level on
parameters box.
Significance level for DModX and Hotellings T2
The significance level used when computing the critical limits for DModX and the
Hotelling's T
2
ellipse is 0.05 (95% confidence).
To use a different significance level than 0.05, type a value between 0 and 1 (not
exactly 0 or 1 but between) in the Significance level for DModX and Hotelling's T2
field.
To not display the limits select None in the Limits page in Properties. For details, see
the Limits subsection in the Plot and list contextual tabs chapter.
Home
179
Reset All button
To reset the options of the Model, Distance to Model, Coefficients, Residuals/R2,
and Predictions pages to the project options, click the Reset All button.
Distance to Model page in Model Options
The Distance to Model page displays:
- Normalized in units of standard deviation.
- Absolute.
- Weighted by the modeling power.

Normalized or absolute distances
The distance to the model (DModX and DModY) can be expressed as an Absolute or a
Normalized value, where the normalized is in units of standard deviation of the pooled
RSD of the model.
By default, the Normalized in units of standard deviation option is selected.
To change from the default, click Absolute.
Weighted by the modeling power
When computing the distance to the model (DModX), the residuals are by default not
weighted by the modeling power of the variables.
To change from the default, select the Weighted by the modeling power check box.
For details, see the MPow weighted residuals DModX section in Appendix A:
Statistics.
Coefficients page in Model Options
The Coefficients page displays the available types of coefficients: Scaled &
Centered, MLR, Unscaled, and Rotated and the Resolve coefficients for
hierarchical top level models check box.

Scaled and Centered
With Scaled & Centered selected, the coefficients displayed are for scaled and
centered X and scaled Y (as specified in the workset). This is the SIMCA default. Use
these coefficients to interpret the influence of the x-variables on Y. Coefficients of
different responses (y-variables) are also comparable as the y-variables are normalized
(scaled).
User Guide to SIMCA
180
MLR
With MLR selected, the coefficients displayed are the PLS coefficients when:
- Y is unscaled and uncentered,
- X is scaled and centered, and
- the second centering and scaling of the cross terms and squares has been
removed.
Unscaled
With Unscaled selected, the coefficients displayed are the coefficients when X and Y
are unscaled and uncentered.
Rotated
With Rotated selected, the coefficients are rotated so they correspond as much as
possible to the pure spectral or other profiles. For more see the Coefficients rotated -
CoeffRot subsection in Appendix A: Statistics.
Resolve coefficients for hierarchical top level models
To automatically resolve coefficients of hierarchical top level models, select the
Resolve coefficients for hierarchical top level models check box.
Note: For hierarchical models, Resolve Coefficients is available from the
Layout tab when the Coefficient Plot or Coefficient Overview Plot is active.
Residuals / R2 page in Model Options
The Residuals / R2 page displays the available types of residuals and R2.

Residuals
The Standardized residuals are displayed by default. The standardized residuals are
the unscaled residuals divided by their standard deviation.
To display the residuals in original, unscaled units, click Raw original units.
R
2

R2 explained variation is selected by default. R
2
is the fraction of the Sum of
Squares (SS) explained by the model.
R
2
adjusted is the fraction of variance explained by the model (SS adjusted for the
degrees of freedom).
To display the adjusted R
2
, click R2 Adjusted variance.
Home
181
Predictions page in Model Options
The Predictions page displays the following check boxes:
- Transform predictions.
- Scale predictions.
- Trim predictions as the workset.

Transform predictions
When the y-variables have been transformed, by default the predictions are back
transformed to the original units.
To display the y-variables in transformed units, select the Transform predictions
check box.
Scale predictions
When the y-variables have been scaled, by default the predictions are rescaled to the
original units.
To display the y-variables in the same units as the workset, select the Scale
predictions check box.
Trim predictions as the workset
When the workset has been trimmed or Winsorized, the predictionset can be trimmed
or Winsorized in the same manner by selecting the Trim predictions as the workset
check box.
By default the predictionset is not trimmed or Winsorized.
Note: When selecting the Trim predictions as the workset check box, all
predictions found on the Predict tab are displayed trimmed.
CV-Groups page
The default is to assign every Nth observation to the same group, where the default N
is defined in the Fit page in Project Options. To change from the default, use the
Model Options, CV-Groups page to specify the assignment of cross validation
groups.
To access the CV-Groups page, the model has to be unfitted. For fitted models, this
page can be opened but all functionality is unavailable.
There are four sections in the CV-Groups page:
- Number of cross validation groups.
- What should the assignment of cross validation groups be based on?
- How should observations be grouped using the selected data above?
- CV Groups list.
User Guide to SIMCA
182
All points are described in the subsections that follow.
Note: Clicking Apply updates the cross validation groups. Neglecting to click
Apply before clicking OK will default back to the original cross validation
groups.

Number of cross validation groups
The number of cross validation groups decides the size of the excluded observation
group during component computation. With the default number of cross validation
groups (7) a 7
th
of the observations is excluded during each cross validation round.
To change from the default, type a number in the Number of cross validation groups
field.
Note: When clicking Apply with the Group observations with the same value
in the same group check box selected, the Number of cross validation groups
field is automatically changed to the number of unique entries.
Cross validation group assignment
There are 4 types of assignments available under What should the assignment of
cross validation groups be based on and 3 types of grouping options under How
should observations be grouped using the selected data above.
In this table the options found under What should the assignment of cross validation
groups be based on? are described:
In this table the options found under How should observations be grouped using the
selected data above? are described:
Option Description of assignment
Group similar observations
in the same group
Groups observations with values numerically close by
first sorting the selected vector (score, variable, or
observation secondary ID), and then parting the
vector in the number of cross validation groups and
setting the first part as CV group 1, the second part
as CV group 2 etc.
Home
183
Option Description of assignment
Group dissimilar
observations in the same
group
Groups observations with values numerically distant
by sorting the selected vector and assigning groups
as in Assign every Nth observation to the same
group (default). For details, see the first row in this
table.
Group observations with the
same value in the same
group
Groups observations with exactly the same value or
text by sorting the selected vector, counting the
number of unique entries, creating CV groups, one
for each unique entry, and then assigning the
observations to the respective groups.
Note that when clicking Apply with this option, the
Number of cross validation groups is automatically
changed to the number of unique entries.
Note: Configuring cross validation groups must be done before the model is
fitted.
Cross validation group examples
In this table the available combinations, with examples, are described:
No. Grouping type Example
Assign every Nth observation to the same group (default)
1. No grouping type
available.
With 7 cross validation groups, the first cross
validation group will hold observations number 7,
14, 21 etc. The second CV group will hold
observations 1, 8, 15, the third 2, 9, 16 etc.
Assign observations based on the score from model Mxx
2. Group similar
observations in the
same group
With 70 observations sorted according to the
value in t[1], and 7 cross validation groups, the
10 observations with the smallest score values
are assigned to CV group 1, etc.
3. Group dissimilar
observations in the
same group
With 70 observations sorted according to the
value in t[1], and 7 cross validation groups, the
first CV group will hold observations with sort
order number 7, 14, 21 etc. The second CV
group will hold observations 1, 8, 15, the third 2,
9, 16 etc. as in 1.
4. Group observations
with the same value in
the same group
The observations are sorted according to the
value in t[1]. CV groups are created, one for each
unique value. Generally this combination of
options will lead to the leave-one-out solution.
Assign observations based on variable V1
User Guide to SIMCA
184
No. Grouping type Example
5. Group similar
observations in the
same group
With 70 observations sorted according to V1, and
7 cross validation groups, the 10 observations
with the smallest values are assigned to CV
group 1, etc.
6. Group dissimilar
observations in the
same group
With 70 observations sorted according V1, and 7
cross validation groups, the first CV group will
hold observations with sort order number 7, 14,
21 etc. The second CV group will hold
observations 1, 8, 15, the third 2, 9, 16 etc. as in
1.
7. Group observations
with the same value in
the same group
The observations are sorted according to V1. CV
groups are created, one for each unique value.
Assign observations based on observation ID
8. Group similar
observations in the
same group
The selected secondary observation ID is sorted.
Then the entire vector is parted in the entered
number of groups, for instance 7. This means
that, with 70 observations and 7 cross validation
groups, the 10 observations with the lowest ID
values/text values are assigned to CV group 1,
etc.
9. Group dissimilar
observations in the
same group
With 70 observations sorted according to the
selected observation ID, and 7 cross validation
groups, the first CV group will hold observations
with sort order number 7, 14, 21 etc. The second
CV group will hold observations 1, 8, 15, the third
2, 9, 16 etc. as in 1.
10. Group observations
with the same value in
the same group
The observations are sorted according to the
selected observation ID. CV groups are created,
one for each unique value.
Cross validation group list
Under CV Groups all included observations are listed in the order of the dataset.
The No column lists the internal numbering, Name lists the primary observation ID,
and Group lists the current cv group assignment.
Clicking the Apply-button updates the CV Groups list Group column according to
the selected options.

Note: Configuring cross validation group must be done before the model is
fitted.
Home
185
More Options
The More Options page contains advanced options that should generally not be
altered. The page displays and allows change of the following:
- T2 Center - used when calculating T2Range and should only be changed to
'0' when planning to import the model in MODDE. In all other cases it
should remain at -99 to use the calculated center.
- Eigenvalue similarity level - is used in the autofit rules for PC modeling.
The default is 0.05 and should generally not be changed.

User Guide to SIMCA
186
Fit model
You can fit the model using Autofit, Two First components, Add component, and
Remove component in the Fit model group.

Autofit
SIMCA extracts as many components as considered significant when clicking Autofit
in the Fit model group, on the Home tab.
When the active model is a wrapper model, BEM or CM, clicking Autofit opens the
Specify Autofit dialog allowing you to autofit all models in the wrapper.
Autofit is only available when using cross validation.
Specifying how to fit the class or phase models
In the Specify Autofit dialog the list displays all class/phase models, all unfitted
models default selected, and Autofit is the planned fit method. For fitted models, the
current number of components is listed under Components and the models are not
selected.
- To not fit a model, clear the corresponding check box by clicking it or
selecting the row and clearing the Include check box.
- To fit a certain number of components, select the corresponding rows, enter
the wanted number in the No. of Components field and click the Set button.
- To autofit a model that has already been fitted, select the model and click the
Autofit button.
Clicking OK closes the dialog and all the selected class models are fitted as specified.

Home
187
Component significant according to cross validation rules
For regular (non-batch) PLS models, a component is significant if it is cross validated
according to rule 1 or rule 2.
For PC models, a component is significant if it is cross validated according to rules 1, 2
or 3.
For OPLS model, a component is significant if it is cross validated according to rule 1
or rule 2.
For O2PLS models, a component is significant if it is cross validated according to rules
1, 2 or 3.
For details about the cross validation rules, see the Cross validation section in
Appendix A: Statistics.
Two First Components
To compute the two first components of the model, click Two First in the Fit model
group, on the Home tab.
Two First is unavailable:
- after the two first components have been computed.
- for OPLS/O2PLS models with more than one y-variable.
Note: With a wrapper model marked, clicking Two First adds the two first
components to all models in the wrapper.
Add component
To compute the next model component, on the Home tab, in the Fit model group,
click Add.
The component, irrespective of its significance, is added to the model if possible.
For OPLS and O2PLS models with more than one y-variable you can select which type
of component to add from the respective galleries. For OPLS the O2PLS specific PCA
orthogonal components are unavailable.

Note: With a wrapper model marked, clicking Add adds the next component to
all models in the wrapper.
User Guide to SIMCA
188
Remove component
When the model has been overfitted, removing components one by one is available by
clicking Remove in the Fit model group on the Home tab.
Remove is available when the model has one or more components.
For OPLS and O2PLS models with more than one y-variable you can select which type
of component to remove from the respective galleries.

Note: With a wrapper model marked, clicking Remove deletes the last
component for all models in the wrapper.
Diagnostics & Interpretation
The Diagnostics & Interpretation group holds the plots and lists commonly used
after fitting a model.
Overview button in the Diagnostics & Interpretation
group
The first button in the Diagnostics & Interpretations group is the Overview button.
When clicked it opens the following plots and tiles them:
- X/Y Overview
- Score Scatter
- Loading Scatter
- DModX
For BLM the following plots are displayed:
- X/Y Overview
- Scores BCC
- Loadings Sources of Variation Plot
- DModX BCC
Home
189
Summary of Fit
The summary plots and lists are found by clicking the Summary of Fit button arrow.
The available plots and lists are: Summary of Fit, X/Y Overview, X/Y Component,
OPLS/O2PLS Overview, Component Contribution, Summary of Fit List, and X/Y
Overview List.

The summary plots by default display R2. To display R2 adjusted instead, right-
click the plot, click Properties, click the R2 tab, and change as desired.
For more about the Properties dialog, see the Tools tab section in the Plot and
list contextual tabs chapter.
Note: For OPLS and O2PLS models, the summary plots and lists only display
statistics for the predictive components for models with more than one
predictive component.
For model statistics for the orthogonal components, see the Model window for
OPLS and O2PLS models subsection earlier in this chapter.
Summary of Fit plot
The Summary of Fit plot displays the cumulative R
2
and Q
2
for the X-matrix for
PCA and for the Y-matrix for PLS/OPLS/O2PLS, after each extracted component.

User Guide to SIMCA
190
Summary of Fit plot for OPLS
For an OPLS or O2PLS model with more than one predictive component, the
Summary of Fit plot displays the cumulative R2 and Q2 for the Y-matrix modeled by
X. For OLS and O2PLS models the components are labeled P, for predictive.
Figure 2. Summary of fit plot showing the progression of R2 and Q2 for the Y-matrix
modeled by X.

For a model with only one predictive component, the Summary of Fit plot displays
how the R2 and Q2 for the Y-matrix evolve (progress) as 1+0, 1+1, 1+2 including the
predictive component and each added orthogonal component to the overall explanation
of the Y-variance.

Home
191
X/Y Overview plot
The X/Y Overview plot displays the individual cumulative R
2
and Q
2
for every X
variable with a PCA model and every Y variable with a PLS model.

X/Y Component plot
The X/Y Component plot displays the R
2
X or R
2
Y and Q
2
X or Q
2
Y cumulative,
per component for the selected X variable with a PCA model or Y variable for a
PLS model.
To switch which variable is displayed, on the Tools tab, in the Data group, use the
Y Variable box or use Properties. For more about the Properties dialog, see the
Plot and list contextual tabs chapter.

User Guide to SIMCA
192
OPLS/O2PLS Overview plot
The OPLS/O2PLS Overview is a newly developed plot, which provides a graphical
summary of the predictive and orthogonal sources of variation in the model. Being an
extension of the conventional Summary of Fit plot of SIMCA, this column plot is
both stacked and color coded in accordance with the predictive and orthogonal
variance structures (Figure 3).
Figure 3. The OPLS/O2PLS Overview plot.

The green color corresponds to the joint X/Y variation. The blue color indicates
orthogonal variations in X or Y respectively, estimated in terms of the OPLS/O2PLS
algorithm. The red color represents orthogonal variations in X or Y, estimated by PCA
subsequent to the convergence of the OPLS/O2PLS algorithm.
Component Contribution plot
The Component Contribution plot for every fitted X, with a PCA model, or Y,
with a PLS model, the plot displays R
2
X or R
2
Y and Q
2
for a selected component.
The selected component number is displayed in the legend.
To change component, on the Tools tab, in the Data group, click the Component
box. Alternatively open Properties, and click the Component tab. For more about
the Properties dialog, see the Tools section in the Plot and list contextual tabs
chapter.

Home
193
Summary of Fit List
The Summary of Fit List displays, for each component and for the total X (PCA) or Y
(PLS/OPLS/O2PLS) matrix:
- R2.
- R2 cumulative.
- Q2 Significant components are colored green.
- Q2 limit.
- Q2 cumulative.
An example for a PCA model:

An example for a PLS model:

For details about the displayed vectors, see the tables in the X/Y Overview List
section next.
Note: For OPLS models with one predictive component this list displays the
progression vectors.
User Guide to SIMCA
194
X/Y Overview List
The X/Y Overview List displays a number of vectors. For PCA models the vectors are
specific for the X block and for PLS/OPLS/O2PLS models for the Y block.
X/Y Overview List for PCA models
For PCA models the X/Y Overview List displays the following for each X variable:
Vector Description
Q
2
VX(Cum) The cumulative predicted fraction (cross-validation) of the variation
of X.
Stdev(X) Standard deviation of X in original units, and transformed if X is
transformed.
RSD(X) Residual standard deviation of X in original units (transformed if the
variable is transformed)
Stdev(X)WS Standard deviation of X scaled as specified in the workset (normally
scaled to unit variance, UV).
RSD(X)WS Residual Standard Deviation, scaled as specified in the workset
(normally scaled to unit variance, UV).
DFR Degrees of freedom of the X residuals.
N Non missing observations in X.
An example for a PCA model:


X/Y Overview List for PLS/OPLS/O2PLS models
For PLS models the X/Y Overview List displays the following for each y-variable:
Vector Description
R
2
Y The cumulative fraction of the variation of the y-variable
explained after the selected component.
R
2
YAdj The cumulative fraction of the variation of the y-variable,
adjusted for degrees of freedom, explained after the selected
component (Cumulative Variance explained).
Q
2
Y (Cum) The cumulative predicted fraction (cross validation) of the
variation of Y.
Stdev (Y) Standard deviation of Y in original units, and transformed if Y
is transformed.
Home
195
Vector Description
RSD (Y) Residual standard deviation of Y in original units (transformed
if the variable was transformed).
Stdev(Y)WS Standard deviation of Y scaled as specified in the workset
(normally scaled to unit variance, UV).
RSD(Y)WS Residual standard deviation scaled as specified in the
workset (normally scaled to unit variance, UV).
DFR Degrees of freedom of the Y residuals.
N Non missing observations in Y.
An example for a PLS model:

Note: For OPLS/O2PLS models, the X/Y Overview List only displays
statistics for the predictive components.
Scores
The tt and uu plots of the X- and Y-scores of, for example, dimensions 1 and 2 (i.e. t
1

vs. t
2
, and u
1
vs. u
2
), can be interpreted as windows into the X- and Y-spaces,
respectively. These show how the X space (X conditions) and response values are
situated with respect to each other. These plots show the possible presence of
outliers, groups, and other patterns in the data.
There are four types of score plots available from the Scores gallery: Scatter, Line,
Column, and 3D.

The tu plots (t
1
vs. u
1
, t
2
vs. u
2
, etc.) display the relation between X and Y. The degree
of fit (good fit corresponds to small scatter around the straight line), indications of
curvature, and outliers can also be seen.
The tu plots are found on the Analyze tab, in the Analysis group, by clicking Inner
Relation.
User Guide to SIMCA
196
Score Scatter Plot
To display the scores in a 2D scatter plot, on the Home tab, in the Diagnostics &
Interpretation group, click Scores.
The default scatter plot is displayed, t1 vs. t2 (PCA, PLS) or t1 vs. to1 (OPLS with
1+1 for instance).
When displaying a two-dimensional score plot displaying t, SIMCA draws the
confidence ellipse based on Hotelling's T
2
, by default at significance level 0.05.
Observations situated outside the ellipse are considered to be outliers. For
example, in the default score scatter plot displayed below it is evident that
observation 208 is an outlier.

Ellipse modifications
To hide the ellipse, or change the confidence level of the ellipse, click the Limits
tab in the Properties dialog and change as desired. For more see the Limits
subsection in the Tools tab section in the Plot and list contextual tabs chapter.
To change the default confidence level use Model Options or Project Options.
For more about the ellipse, see the Hotellings T2 subsection in Appendix A:
Statistics.
Color by
To color by a variable, or scores etc, click the Color by button in the Tools tab or
the Color tab in the Properties dialog. For more see the Coloring from Properties
subsection in the Tools tab section in the Plot and list contextual tabs chapter.
Home
197
Score plot Properties

In the Scores page of the Properties dialog you can select to:
- Display more or other series: make the selection in Item and Comp
boxes, and click Add Series.
- Modify the items on the axes.
- Clear the Scale proportionally to R2X check box for OPLS and O2PLS
if warranted. For more about this option, see the R2X subsection in the
Plot/List chapter.
The other tabs are general and described in detail in the Properties dialog section
in the Tools section in the Plot and list contextual tabs chapter.
Note: When the dataset contains numerical IDs, they can be selected and
plotted on either axis.
Removing ellipse in score plot
There are two ways of removing the ellipse in the Score Scatter Plot:
- Not display it by opening Format Plot, under the Limits and Regions node,
click Ellipse and select Line Style None and Fill Type No fill.
- Hide it in the Limits page in the Properties dialog. For more, see the Limits
subsection in the Tools tab section in the Plot and list contextual tabs
chapter.
User Guide to SIMCA
198
Score Line Plot
To display the scores in a line plot, on the Home tab, in the Diagnostics &
Interpretation group, click Scores, and then click Line.
The default line plot of t1 is displayed.

With a one-dimensional score line plot, e.g. t
a
vs. Num, the score plot displays
confidence intervals corresponding to the 2 and 3 sigma limits, i.e., 2 and 3
standard deviations of the vector are displayed by default. Here the plot is
identical to the Shewhart control chart with no subgroups.

In the Scores page of the Properties dialog you can select to:
- Display more or other series: make the selection in Item and Comp
boxes, and click Add Series.
- Modify the items on the axes.
- Clear the Scale proportionally to R2X check box for OPLS and O2PLS
if warranted. For more about this option, see the R2X subsection in the
Plot/List chapter.
The other tabs are general and described in detail in the Properties dialog section
in the Tools tab section in the Plot and list contextual tabs chapter.
Home
199
Note: When the dataset contains numerical IDs, they can be selected and
plotted on either axis.
Score line plot for batch evolution models
Displaying the scores as a line plot in combination with DModX allows you to detect
batches with upsets. These plots use the non-aligned batches.
The limits are computed from the variation of the ts as single vectors, as for steady
state data. Batches outside the limits or different from the average behavior should be
examined further, for example by displaying contribution plots. These batches should
be removed from the workset if they are outlier batches.

See also the DModX plot for BEM subsection later in this chapter.
The significance level can be changed by clicking the Limits tab in the Properties
dialog.
User Guide to SIMCA
200
Score Column Plot
To display the scores in a column plot, on the Home tab, in the Diagnostics &
Interpretation group, click Scores, and then click Column.
The default column plot of t1 is displayed.

With a column plot displaying one vector, jack-knifing is used to calculate
standard errors displayed as error bars at the end of each column.

In the Scores page of the Properties dialog you can select to:
- Display more or other series: make the selection in Item and Comp
boxes, and click Add Series.
- Modify the items on the axes.
- Clear the Scale proportionally to R2X check box for OPLS and O2PLS
if warranted. For more about this option, see the R2X subsection in the
Plot/List chapter.
Jack-knifing is used to calculate standard errors displayed as error bars at the end
of each column. The confidence level of the standard errors can be customized in
the Limits page.
The other tabs are general and described in detail in the Properties dialog section
in the Tools tab section in the Plot and list contextual tabs chapter.
Home
201
Score Scatter 3D Plot
To display the scores in a 3D scatter plot, on the Home tab, in the Diagnostics &
Interpretation group, click Scores, and then click 3D.
The default scatter plot is displayed, t1 vs. t2 vs. t3 (PCA, PLS) or t1 vs. to1 vs to2
(OPLS with 1+2 for instance).


In the Scores page of the Properties dialog you can select to:
- Display more or other series: make the selection in Item and Comp
boxes, and click Add Series.
- Modify the items on the axes.
- Clear the Scale proportionally to R2X check box for OPLS and O2PLS
if warranted. For more about this option, see the R2X subsection in the
Plot/List chapter.
The other tabs are general and described in detail in the Properties dialog section
in the Tools tab section in the Plot and list contextual tabs chapter.
Note: When the dataset contains numerical IDs, they can be selected and
plotted on any axis.
User Guide to SIMCA
202
Marking in 3D scatter plot
In the 3D plot, hold down the CTRL-button to mark several points.
To mark two groups, in order to create a group contribution plot, mark the first group
then click a point after releasing the CTRL-button (so that the first group is Previously
Marked Values) then mark the second group while pressing the CTRL-button. Open
the Marked Items window to see the observations included in each group.
Note: Avoid marking the single point again as it will then be unmarked.
Zooming in 3D plot
To zoom, scroll using the mouse wheel. Alternatively, press the CTRL-button while
holding down the right mouse button and moving the mouse.
Rotating 3D plot
To rotate a 3D scatter plot or a Response Surface plot, hold down the left mouse
button and move the mouse in the direction to turn it. To keep the plot rotating, release
the mouse button while turning. To not have the plot continue rotating, stop turning
before releasing the mouse button.
Resetting the rotation
To reset the rotation to default, right-click the plot and then click Reset rotation.
Moving the 3D plot in its window
To move the plot, hold down the right mouse button and move the mouse in the
direction to move. Moving the plot is useful after having zoomed when wanting to see
other areas of the plot.

Home
203
OPLS specific score plots
There are four OPLS/O2PLS specific score plots, each displaying different
components of the model.
If there are two or more of the score type in the model, the plots are scatter plots of the
first two components. If there is only one component of the type available, the
resulting plot is a column plot. By default the plotted score vectors are scaled
proportionally to R2X, the variance explained by each component.
The OPLS/O2PLS specific score plots are:
- Pred X X-score vectors t1 and t2 of the predictive components.
- Pred Y Y-score vectors u1 and u2 of the predictive components.
- Orth X X-score vectors to1 and to2 of the orthogonal in X components.
- Orth Y Y-score vectors uo1 and uo2 of the orthogonal in Y components.
User Guide to SIMCA
204
Loadings
Loading plots display the correlation structure of the variables. That is, they show
the importance of the x-variables in the approximation of the X matrix.
There are four types of loading plots: Scatter, Line, Column, or 3D.

Note: The score and loading plots complement each other. The position of
objects in a given direction in a score plot is influenced by variables lying in
the same direction in the loading plot.
The available loading vectors for PCA, PLS, OPLS and O2PLS models are
described in the table in the Observations and Loadings vectors subsection in
Appendix A: Statistics.
Loading
vector
Short description
c PLS only.
Displays the correlation between the Y variables and the X scores
T(X).
Weights that combine the Y variables with the scores u, so as to
maximize their correlation with X.
Y variables with large c's are highly correlated with T(X).
p Displays the importance of a variable in approximating X as TP.
pc

PLS only.
Combination of the p and c vectors.
c(corr) PLS only.
Available after selecting Correlation Scaled in the loading dialog.
c scaled as correlations resulting in all points falling inside the circle
with radius 1.
p(corr) Available after selecting Correlation Scaled in the loading dialog.
p scaled as correlations resulting in all points falling inside the circle
with radius 1.
pc(corr) PLS only.
Available after selecting Correlation Scaled in the loading dialog.
Combination of the p(corr) and c(corr) vectors.
w PLS only.
Weights that combine the X variables (first dimension) or the residuals
of the X variables (subsequent dimensions) to form the scores t.
Home
205
Loading
vector
Short description
w* PLS only.
Weights that combine the original X variables (not their residuals as
with w) to form the scores t.
w*c PLS only.
Combination of the w* and c vectors.
Plotting both the X-weights, w
*
, and Y-weights, c, in the same plot
displays the correlation structure between X and Y. Interpretation of
how the X and Y variables combine in the projections, and how the X
variables relate to the Y variables, is possible.
Loading Scatter Plot
To display the loadings in a 2D scatter plot, on the Home tab, in the Diagnostics
& Interpretation click Loadings, and then click Scatter.
The default scatter plot is displayed, p1 vs. p2 (PCA), w*c1 vs. w*c2 (PLS) or
pq1 vs. poso1 (OPLS with 1+1 for instance).
To change what is displayed, open the Properties dialog by clicking the plot and
then Properties in the mini-toolbar.

In the Loadings page of the Properties dialog you can select to:
- Display more or other series: make the selection in Item and Comp
boxes, and click Add Series.
- Modify the items on the axes.
- Display loadings vectors p, c, or pc scaled as correlations: select the
Correlation scaled check box. The correlation scaling done is the same
as in the Biplot. For more, see the Biplot subsection later in the Analyze
chapter.
- Clear the Normalize to unit length check box. Normalize to unit length
check box is default selected for OPLS and O2PLS. For more about this
transformation, see the Normalize subsection in the Plot/List chapter.
The other tabs are general and described in detail in the Properties dialog section
in the Tools tab section in the Plot and list contextual tabs chapter.
User Guide to SIMCA
206
Note: When the dataset contains numerical IDs, they can be selected and
plotted on either axis.
Loading Scatter Plot PCA
Variables with the largest absolute values of p1 or/and p2 dominate the projection.
Variables near each other are positively correlated; variables opposite to each other are
negatively correlated.

Loading Scatter Plot PLS
Variables with large w* or c values are situated far away from the origin (on the
positive or negative side) in the plot.
X variables with large values in w* (positive or negative) are highly correlated with U
(Y).

The default coloring of the loading scatter plot is by Terms.
Home
207
Loading Line Plot
To display the loadings in a line plot, on the Home tab, in the Diagnostics &
Interpretation group, click Loadings, and then click Line.
The default line plot is displayed, p1 (PCA), w*c1 (PLS) or pq1 (OPLS).
To change what is displayed, open the Properties dialog by clicking the plot and
then Properties in the mini-toolbar.
The following dialog opens:

In the Loadings page of the Properties dialog you can select to:
- Display more or other series: make the selection in Item and Comp boxes,
and click Add Series.
- Modify the items on the axes.
- Display loadings vectors p, c, pc, pq, po, poso, scaled as correlations: select
the Correlation scaled check box. The correlation scaling done is the same
as in the Biplot. For more, see the Biplot subsection later in the Analyze
chapter.
- Clear the Normalize to unit length check box. Normalize to unit length
check box is default selected for OPLS and O2PLS. For more about this
transformation, see the Normalize subsection in the Plot/List chapter.
The other tabs are general and described in detail in the Properties dialog section
in the Tools tab section in the Plot and list contextual tabs chapter.
Note: When the dataset contains numerical IDs, they can be selected and
plotted on either axis.
User Guide to SIMCA
208
Loading Line Plot PCA
Variables with the largest absolute values of p1 dominate the projection.

Loading Line Plot PLS
Variables with large w* or c values are situated far away from the origin (on the
positive or negative side) in the plot.
X variables with large values in w* (positive or negative) are highly correlated
with U (Y).

Home
209
Loading Column Plot
To display the loadings in a column plot, on the Home tab, in the Diagnostics &
Interpretation group, click Loadings and then Column.
The default column plot is displayed, p1 (PCA), w*c1 (PLS) or pq1 (OPLS).
To change what is displayed, open the Properties dialog by clicking the plot and
then Properties in the mini-toolbar.

In the Loadings page of the Properties dialog you can select to:
- Display more or other series: make the selection in Item and Comp boxes,
and click Add Series.
- Modify the items on the axes.
- Display loadings vectors p, c, pc, pq, po, poso, scaled as correlations: select
the Correlation scaled check box. The correlation scaling done is the same
as in the Biplot. For more, see the Biplot subsection later in the Analyze
chapter.
- Clear the Normalize to unit length check box. Normalize to unit length
check box is default selected for OPLS and O2PLS. For more about this
transformation, see the Normalize subsection in the Plot/List chapter.
With a column plot, jack-knifing is used to calculate standard errors displayed as error
bars at the end of each column. The confidence level of the standard errors can be
customized in the Limits page.
The other tabs are general and described in detail in the Properties dialog section in
the Tools tab section in the Plot and list contextual tabs chapter.
User Guide to SIMCA
210
Loading Column Plot PCA
Variables with the largest absolute values of p1 dominate the projection.

Loading Column Plot PLS
Variables with large w* or c values are situated far away from the origin (on the
positive or negative side) in the plot.
X variables with large values in w* (positive or negative) are highly correlated
with U (Y).

Home
211
Loading Scatter 3D Plot
To display the loadings in a 3D scatter plot, on the Home tab, in the Diagnostics
& Interpretation group, click Loadings, and then click 3D.
The default 3D scatter plot is displayed, p1 vs. p2 vs. p3 (PCA), w*c1 vs. w*c2
vs. w*c3 (PLS) or pq1 vs. poso1 vs. poso2 (OPLS with 1+2 components for
instance).
To change what is displayed, open the Properties dialog by clicking the plot and
then Properties in the mini-toolbar.

In the Loadings page of the Properties dialog you can select to:
- Display more or other series: make the selection in Item and Comp
boxes, and click Add Series.
- Modify the items on the axes.
- Display loadings vectors p, c, pc, pq, po, poso, scaled as correlations:
select the Correlation scaled check box. The correlation scaling done is
the same as in the Biplot. For more, see the Biplot subsection later in the
Analyze chapter.
- Clear the Normalize to unit length check box. Normalize to unit length
check box is default selected for OPLS and O2PLS. For more about this
transformation, see the Normalize subsection in the Plot/List chapter.
The other tabs are general and described in detail in the Properties dialog section
in the Tools tab section in the Plot and list contextual tabs chapter.
Note: When the dataset contains numerical IDs, they can be selected and
plotted on any axis.
User Guide to SIMCA
212
Loading Scatter 3D Plot example
Variables with the largest absolute loading values dominate the projection. Variables
near each other are positively correlated; variables opposite to each other are
negatively correlated.

For how to mark, zoom, rotate, move the plot in its window, color and size the 3D plot,
see the Score Scatter 3D Plot section earlier in this chapter.
OPLS specific loadings
There are three OPLS and O2PLS specific loading plots, each displaying different
components of the model.
If there are two or more components for the loading type to display, the displayed plot
is a scatter plot. If there is only one component, the resulting plot is a column plot.
By default the plotted loading vectors are scaled to unit length.
- Pred X-Y: X-Y-loading vectors pq1 and pq2 of the predictive components.
- Orth X: X-loading vectors poso1 and poso2 of the orthogonal in X
components.
- Orth Y: Y-loading vectors qor1 and qor2 of the orthogonal in Y
components.
Home
213
Hotellings T2Range
The Hotelling's T2Range can be displayed as two plot types, Line and Column, by
clicking Hotelling's T2 in the Diagnostics & Interpretation group on the Home tab.
The Hotelling's T
2
Range plot displays the distance from the origin in the model plane
(score space) for each selected observation. The plot shows the T
2
calculated for the
range of selected components, i.e., 1 to 7, or 3 to 6.
Default is to display from the first to the last component. For OPLS and O2PLS
models the range is locked to from the first predictive to the last orthogonal in X. To
change the range of components, right-click the plot and select Properties, and then
click the Component tab. y

For all other pages in Properties, see the Properties dialog subsection in the
Tools tab section in the Plot and list contextual tabs chapter.
The Hotelling's T2 subsection in Appendix A: Statistics describes how the
Hotelling's T2 is calculated.
Limits in Hotellings T2Range
Values larger than the yellow limit are suspect (0.05 level), and values larger than the
red limit (0.01 level) can be considered serious outliers.
A large T
2
-range value for a given observation, i.e., a value far above the critical limits,
indicates that the observation is far from the other observations in the selected range of
components in the score-space. Hence, this is likely to be an outlying observation that,
if included the workset, may pull the model in a detrimental way.
User Guide to SIMCA
214
Distance to Model
Distance to Model is an estimate of how far from the model plane, in the X or Y
space, the observation is positioned.
The distance to the model can be displayed in absolute and normalized units. By
default the distance to model plot is displayed in normalized units after the last
component with limit for significance level 0.05. To change from the defaults, use
Model Options or Project Options. The significance level can be changed in the
Limits page in the Properties dialog. For more, see the Limits subsection in the Tools
tab section in the Plot and list contextual tabs chapter.
There are two plots displaying the distance to the model, DModX and DModY,
available on the Home tab in the DModX gallery:
- Under the DModX header, click Line or Column to display DModX in that
plot type.
- Under the DModY header, click Line or Column to display DModY in that
plot type.
Overview of calculation of distance to the model
The RSD of an observation in the X or Y space, is proportional to the observation
distance to the hyper plane of the model in the X or Y space. SIMCA computes the
observation distances to the model, in the X space (DModX) or Y space (DModY).
DModX can also be weighted by the modeling power by selecting the Weighted by
the modeling power check box in the Distance to Model page in Model Options or
Project Options.
For details about the calculations of the distance to the model, see the Distance to
Model section in Appendix A: Statistics.
Distance to the model in the X-block
The distance to the model in the X space, DModX, is by default displayed in
normalized units.
Larger DModX than the critical limit indicates that the observation is an outlier in the
X space.

Home
215
Distance to the model in the Y-block
The distance to the model in the Y space, DModY, is by default displayed in
normalized units.
A large DModY value indicates that the observation is an outlier in the Y space.

DModX plot for batch evolution models
The D-Crit is computed as for steady state data and from an F distribution. Hence it is
very sensitive to tails. Batch data are usually dynamic, and some variables increase or
decrease with time. Hence, batches will often be outside the DModX limit at the
beginning or the end of the evolution. To detect abnormal batches, look for batches far
above the average of the high points.

To exclude batches in a plot, first select the batch using Batch marking mode, the
whole batch will be marked, then use the Exclude tool (red arrow), on the Marked
Items tab, to exclude the batch.
When batches have phases, the batch will be excluded from all the class (phase)
models and a new unfitted batch model, with all class models, generated. You can
mark several batches holding the CTRL key, and then exclude them.
See also the Batch Control Charts section later in this chapter.
User Guide to SIMCA
216
Observed vs. Predicted
The Observed vs. Predicted plot, displayed by clicking Observed vs. Predicted in
the Diagnostics & Interpretation group, on the Home tab, displays the observed
values vs. the fitted or predicted values for the selected response, after the last
component.
To switch to another y-variable use the Tools tab (X-Axis YVar box in the Data
group) or the Properties dialog (Select Y-variable tab). For more, see the Switching
components, batches, and models subsection respective the Select Y Variable
subsection in the Tools tab section in the Plot and list contextual tabs chapter.

Coefficients
From the PLS model SIMCA computes PLS regression coefficients for each Y
variable. These express the relation between the Y variables and all the terms in the
model. By default, the regression coefficients relate to the centered and scaled data,
CoeffCS, and are computed from all extracted components.
Note: All coefficients are cumulative.

For PLS/OPLS/O2PLS models two plots and one list are available on the Home
tab by clicking Coefficients:
- Plot - column plot displayed with uncertainty bars for the selected y-
variable.
- Overview Plot - column plot displaying all y-variables.
- List - displayed for all y-variables with uncertainty estimates.
For hierarchical top models, the Resolve Coefficients check box is available in the
Show group on the Layout tab when the Coefficient Plot or Coefficient
Overview Plot is active.
Home
217
Coefficient types
The coefficient plots can be displayed for Scaled and centered, MLR, Unscaled, and
Rotated coefficients. The confidence intervals are only available for coefficients of
scaled and centered data.
Displaying the Rotated coefficients is relevant when the X block is spectral data. The
rotated coefficients express the pure spectra as it relates to each y-variable.
To switch coefficient type, click the plot and in the mini-toolbar click Properties.
Click the Coefficients tab, and then click the desired coefficient type.
For all other pages in Properties, see the Properties dialog subsection in the
Tools tab section in the Plot and list contextual tabs chapter.
For details about the different coefficient types, see the Coefficients subsection in
Appendix A: Statistics.
For hierarchical top models, the Resolve Coefficients check box is available in the
Show group on the Layout tab when the Coefficient Plot or Coefficient
Overview Plot is active.
Coefficient Plot
The Coefficient Plot by default displays the coefficients referring to scaled and
centered data for a given response, with confidence intervals derived from jack-
knifing. The confidence level of these limits can be modified in the Properties
dialog. For more, see the Limits subsection in the Tools tab section in the Plot
and list contextual tabs chapter.

Note: For hierarchical top models, the Resolve Coefficients check box is
available in the Show group on the Layout tab when the Coefficient Plot or
Coefficient Overview Plot is active.
User Guide to SIMCA
218
Use the Properties dialog to switch coefficient type.

Coefficient Plot for hierarchical top level models
In hierarchical top level models, coefficients refer to the upper level variables,
usually scores of the base level models. For the interpretation of the upper level
model, it is desirable to be able to translate the upper level coefficients to
coefficients of the individual variables of the base models.
For hierarchical top models the Coefficient Plot by default displays the top level
coefficients. To resolve to the base model coefficients, on the Layout tab, in the
Show group, select the Resolve Coefficients check box.
To change this default to display the resolved coefficients, see the Coefficients
page in Model Options subsection in the Workset section earlier in this chapter.
Note: Only linear terms are resolved. If the top model contains expanded
terms (interaction, squares etc.) these terms are not resolved.

Home
219
Coefficient Overview Plot
The Coefficient Overview Plot displays the coefficients, referring to scaled and
centered data, for every response as bar graphs side by side.

Coefficient List
The Coefficient List default displays the coefficients referring to scaled and
centered data for all responses, with confidence interval derived from jack-knifing.

User Guide to SIMCA
220
VIP
Interpreting a PLS model with many components and a multitude of responses can be a
complex task. A parameter which summarizes the importance of the X-variables, both
for the X- and Y-models, is called the variable influence on projection, VIP. VIP is a
weighted sum of squares of the PLS weights, w*, taking into account the amount of
explained Y-variance in each dimension. Its attraction lies in its intrinsic parsimony;
for a given model and problem there will always be only one VIP-vector, summarizing
all components and Y-variables. One can compare the VIP of one term to the others.
Terms with large VIP, larger than 1, are the most relevant for explaining Y.
The VIP values reflect the importance of terms in the model both with respect to Y, i.e.
its correlation to all the responses, and with respect to X (the projection). With
designed data, i.e. close to orthogonal X, the VIP values mainly reflect the correlation
of the terms to all the responses. VIP values are computed, by default, from all
extracted components.
The VIP plot is available on the Home tab, in the Diagnostics & Interpretation
group, by clicking VIP.
Note: The VIP plot is cumulative.
VIP Plot
The VIP plot displays the VIP values as a column plot sorted in descending order
with confidence intervals derived from jack-knifing.

To display an unsorted VIP plot (useful in spectroscopy and chromatography), create
the VIP plot in the following manner:
1. On the Plot/List tab, click Column.
2. In the Select data type box select Observations and Loadings.
3. In the Item box select VIP.
4. Click Add Series, and then OK.
The plot is displayed with jack-knife uncertainty bars. The confidence level of these
limits can be modified in the Properties dialog. For more, see the Limits subsection in
the Tools tab section in the Plot and list contextual tabs chapter.

221
Data
Introduction
This chapter describes all commands on the Data tab.
On the Data tab the following commands are available:
1. Merge datasets observation or variable wise.
2. Split the selected dataset variable wise.
3. Transpose dataset making the current columns rows and vice versa.
4. Delete Dataset to remove datasets from the project.
5. Generate Variables from variables in the dataset or from results of fitted
models.
6. Local Centering allowing importing local centering and then viewing it.
7. Spectral Filters: Derivative, MSC, SNV, Row Center, Savitzky-Golay,
EWMA, Wavelet Compression, Wavelet Denoising, OSC, etc. as individual
filters, or chained in any combination.
8. Time Series Filters: Wavelet Compression and Wavelet
Denoising/Decimation of PLS Time series.
9. Dataset Summary displaying the properties of the selected dataset including
Filter tab after filtering.
10. Trimming Overview available after trimming or Winsorizing.
11. Spectra plot displaying XObs for all observations in the selected dataset.
12. Hierarchical base model resulting in hierarchical dataset according to
selection.

Merge
Merging of two datasets is available by clicking Merge in the Modify Dataset group
on the Data tab.
User Guide to SIMCA
222
In this dialog, datasets can only be merged by matching primary IDs. The resulting file
will include the union of the available variables and observations.

To merge:
1. In the First dataset (destination) box select the dataset that should end up
on top or to the left.
2. In the Second dataset (source) box select the dataset that should end up at
the bottom or to the right.
Note: Datasets created by SIMCA (filtered datasets, batch level datasets etc.)
cannot be merged.
Split dataset
Datasets can be split in two by clicking Split in the Modify Dataset group on the Data
tab.

Data
223
In the Split Dataset dialog:
1. Select the Dataset to split.
2. Enter a dataset name in the New dataset field.
3. In the Available Variables mark the variables to split out to the new dataset
and click the >>-button.
4. Click OK to create the new dataset splitting the specified variables from the
original dataset.
Note: Datasets created by SIMCA (filtered datasets, batch level datasets etc.)
cannot be split.
Note: This feature is also available last in the Spectral Filters wizard after
excluding some variables.
Transpose
Transposing of datasets is available by clicking Transpose in the Modify Dataset
group on the Data tab.
When clicking Transpose a message is displayed stating that when transposing the
dataset the following changes happens:
- All dependent datasets will be deleted.
- The current predictionset will be deleted.
- All dependent models will be deleted.
- Qualitative variables will lose their descriptors, that is, will no longer be
qualitative variables but the descriptors will be replaced by integers
according to their order.

Note: Datasets created by SIMCA (filtered datasets, batch level datasets etc.)
cannot be transposed, nor can batch evolution datasets.
User Guide to SIMCA
224
Delete Dataset
Delete datasets by clicking Delete Dataset in the Modify Dataset group on the Data
tab. Any dependent datasets, predictionsets and models are automatically deleted.

Note: The first imported dataset cannot be deleted.
Generate Variables
New variables can be generated by using a function, an expression with operators, or
model results, in the formula field on the first page of the Generate Variables wizard.
Function names are case insensitive and the result of a function can be passed as an
argument to other functions. The result of an expression must yield a matrix with one
or more columns where the columns will become new variables.
Note: To generate new observations, first transpose the dataset, then use the
Generate Variables feature, and finally transpose the dataset again.
In the subsections that follow, the Generate Variables wizard, user generated
functions, creation of new variables from other datasets, and generating variables in
batch projects are covered.
Generate Variables wizard
Creating new variables as functions of existing ones, or from the results of fitted
models, is available on the Data tab, in the Modify Dataset group, by clicking
Generate Variables. All new variables are appended to the right of the last column of
the selected dataset.
Data
225
To add generated variables:
Step Description Illustration
1. Start generating variables by
clicking Generate Variables.

2. Select dataset, in the dialog
listing the available datasets.
When only one dataset is
available, this dialog is not
opened.

3. In the Generate Variables
wizard enter the expression
defining the new variable(s).
Click the question mark button
to read about the operators
and functions available. Click
Next.
Note: Power, multiplication,
and division cannot be applied
to a set of variables and
another set of variables. The
first operand can be a constant,
a variable or a set of variables,
but the second and following
operands must be a single
variable or a constant.

4. The new variables are
displayed with their formula,
statistics and Quick info plots.
By default the new variables
are named according to their
recipe. To change the names,
see point 5, otherwise go to
point 6.

User Guide to SIMCA
226
Step Description Illustration
5. Change the names as desired,
either manually or by clicking
the Change Name button
available when marking the
current variable name. In
Change Variable ID select
Use Recipe (default) or
Specify (name followed by
number). When selecting
Specify (name followed by
number) a sequential number
is automatically added after the
name specified in the Name
field. Click OK.

6. Click Finish to exit the wizard
or Finish Generate More to
reopen the Generate
Variables wizard and continue
at point 3.
When clicking Finish after
adding variables to the dataset,
if there are more datasets,
SIMCA will display a dialog
enabling adding the variables
to other datasets also.

User generated functions
You can generate your own special functions to preprocess the data by creating a DLL
and put it into a directory called Plugins.
In File | SIMCA Options | More Options specify or view the path of the Plugin
Directory.
These functions can be any mathematical formula provided they return a matrix.
For example, the DLL might contain the following functions:
- MATRIX* LowPassFilter(MATRIX* pMatrix, float
fValue);
- MATRIX* HighPassFilter(MATRIX* pMatrix, float
fValue);
- MATRIX* FIR(MATRIX* pMatrix, int iWindowType,
int iFilterLength, ...
To use these functions in SIMCA, simply type f: and the name of the function. The f:
is to differentiate user provided functions from SIMCA functions.
The functions can be nested. For example you may type in Generate Variables:
f:LowPassFilter(f:HighPassFilter(v1, arg1), arg2)
This will apply a high pass filter with arg1 and then a low pass filter with arg2 on
variable number 1 in the dataset.
For detailed information, contact your Umetrics sales office.
Data
227
New variables from other dataset
You can generate new variables from any existing SIMCA datasets.
SIMCA accepts the following syntax:
Dint1: v[int2: int3]
Variables number int2 to int3 in dataset int1.
This expression is only valid if the current dataset has the same number of observations
as dataset number int1
Examples:
D2:v[3,6,8:11]
Generates variables 3,6,8,9,10,11 from dataset 2.
Note: Dataset number 1 is the first dataset. New variables are always
appended at the end of the selected dataset.
Generating variables in batch projects
With batches all functions operating on the whole vector (variable) are implemented
batch by batch within a phase. When batches have phases, the default is to apply the
function to all phases. If you want the function to operate on a variable in only selected
phases, you must use the phase function and specify the desired phases.
The phase specific functions are described in the ?-help in the Generate Variables
wizard in SIMCA, but here follows some specifics.
Note: For batch level datasets, Generate Variables can only add variables to
batch condition datasets, not to the dataset created by SIMCA.
Batch specific syntax
The syntax is as follows:
Function Phase (set-of-variables, integer-set)
Where:
- Function is the selected function to apply.
- Phase signifies that the function should be applied phase wise
- Set-of-variables is any matrix resulting from other functions or a set of
variables.
- integer-set is a set of one or more numbers, representing the phase numbers.
Enclose the numbers with square brackets if you want to specify more than
one phase. For example: [1, 3, 5:10] means phases 1, 3 and 5 to 10.
Applying a function to some phases
When the function operates on some phases only, the default is to leave the values of
the variables as they are for the omitted phases.
If you want to have missing values for the variable in the omitted phases, or a specific
value for example 0, you must add the text misval or the fvalue at the end of the
function.
User Guide to SIMCA
228
For example:
Function Phase (set-of-variables, integer-set, misval) will set the variable to missing
in the omitted phases.
While Function Phase (set-of-variables, integer-set, 0) will set the variable to 0 in the
omitted phases.
Setting the values to 0 in the omitted phases, allows you to apply different functions to
the same variable in different phases,
For example:
Low Pass (Phase (v1, 2, 0)) + HiPass (Phase (v1, [1,3], 0)) will generate a new variable
by applying a Low Pass filter to variable v1 in phase 2 and a Hi Pass filter to variable
v1 in phase 1 and 3.
Generate Variables for batch level without batch conditions
When you have no batch condition datasets but want to generate variables
built on the batch level dataset you first need to create a batch condition
dataset. Create this batch condition dataset as follows:
1. Copy the batch ID of the batch level dataset.
2. File | Import Dataset.
3. Select an empty spreadsheet (in SIMCA import File | New Spreadsheet |
Blank.
4. Paste the batch ID and specify it as batch ID.
5. Type a variable name and leave all cells blank.
6. Specify the first row as primary variable ID:
7. Verify that the Issues pane will exclude all rows and columns that lack batch
ID or primary variable ID and click Resolve All.
8. Click Finish to import the dataset. Now use this dataset to add the generated
variables to by selecting the dataset in the Generate Variables dialog. Note
that you have to use the dataset number to refer to the other dataset. See
dataset numbers in the Plot/List dialogs.
Local Centering
Local centering is useful when applying run-to-run control and changing set points.
With local centering one models the variation of the:
- Batch around its specified center in a phase for batch projects.
- Variable, or part of variable with for instance classes, around the specified
center.
The imported local centering values are applied to all selected datasets when fitting the
model and when making predictions.
Data
229
Importing local centering
To import local centering:
1. On the Data tab, in the Modify Dataset group, click Local Centering |
Import.
2. Select the file containing the specifications for local centering of the
variables.
3. In the dialog, format according to the below:
- Var ID is the primary variable as a row, specifying the variables to
center.
- Center ID is the Secondary Observation ID/Class ID/Batch ID
as a column, specifying the observations to center. Using a
secondary observation ID to specify which observations to center
together can be useful when the dataset has classes. When using
the Batch ID as center ID, the column may hold each batch ID
only once, or once per phase. Specifying batch ID is necessary for
batch evolution datasets where you want to specify the local
centering per batch.
- Phase ID as a column with each phase ID only available once per
batch. Specifying Phase ID is necessary for batch projects with
phases.
- Centers, used for the subtraction, as data in the spreadsheet.

Note: The local centering specification specifies the values to subtract from
the selected variables.
User Guide to SIMCA
230
Clicking Next opens the Local Centering Summary page.

Local centering of the predictionset
The predictionset dataset uses the imported local centering when available.
See also the Local centering missing subsection.
Local centering missing
After local centering has been imported, variables that lack local centering value are
treated as follows:
Batch project
When the local centering was applied according to Batch ID, but no local centering is
available for some batches, the locally centered variables of that batch are locally
centered using the average of the local centering values of all batches included in the
workset.
If you want to import the original values for a certain batch, specify local centering as
'0' for that batch/variable combination.
When local centering is missing for predictions, the average of the workset batches for
that variable is used in the local centering.
Regular project
When the local centering was applied according to Center ID, but no local centering is
available for some classes, the observations are not centered at all; their original value
is used.
View imported local centering
The imported local centering values can be displayed by clicking Local Centering |
View. If View is unavailable, no local centering has been imported.
Note: Local centering is only saved if the variable is available in one of the
imported datasets.
Data
231
Spectral Filters
To apply spectral filters observation wise, on the Data tab, in the Filters group, click
Spectral Filters. The three filter methods Derivatives, MSC and SNV are available
by clicking the arrow.
Clicking Spectral Filters opens the following dialog:

The filters listed in this dialog are described in the subsections that follow. A short
background is available in Appendix B: Preprocessing.
To specify which filter or chain of filters to use:
1. Click the desired filter in the Available list.
2. Click the => button.
3. Repeat 1-2 until the Selected list displays the desired filters.
4. If there are more than one dataset in the project, select the desired dataset in
the Source box.
5. Click OK and the wizard of each filter opens successively and the filters are
applied in the order specified in the Spectral Filters dialog.
6. After completed filtering, a new filtered dataset is created, leaving all
original variables in the original dataset. Optionally excluded variables can
be split out to a separate dataset by selecting the Split out new variables to
a new dataset check box on the last page of the wizard. For more about
splitting datasets, see the Split subsection previously in this chapter.
SIMCA also supports user written filters. For more information, contact your Umetrics
sales office.
Filtering limitations
The following limitations are present for filtering:
- Only one of the wavelet filters can be included in a chain. All other filters
can be applied in any order or combination.
- No filtering is available for filtered converted projects.
- Qualitative variables cannot be filtered and are automatically excluded and
cannot be included.
- Projects with local centering data cannot be filtered.
- Only one dataset at a time can be filtered. To filter two datasets, first Merge
them.
User Guide to SIMCA
232
First, Second, and Third Derivatives
Applying derivatives transforms the dataset from the original domain to the first,
second, or third derivate. For more, see the Derivatives section in Appendix B:
Preprocessing.
When selecting Derivatives in the Spectral Filters dialog and clicking Next, the first
page of the derivatives wizard opens displaying lists of variables and observations.

Selecting the variables and observations to transform
To apply the filter to the dataset, on the first page of the wizard:
1. In the Variables list, mark and Exclude unwanted variables. All variables
specified as X in the default workset are by default included, while those
specified as Y are by default excluded. Qualitative variables cannot be
filtered.
2. In the Observations list, mark and Exclude the observations to exclude from
the filtering. All observations are by default included.
3. To specify variables and observations according to a fitted model, select the
model in the As model box. Variables in the model specified as Y are
automatically excluded.
4. Click Next.
Data
233
Specifying parameters for derivation
Specify the parameters for the derivation in the Derivation specification section by
following the steps described here. Note that marking an observation updates the
graphs displaying the Original data (spectra) and the Derivated data with the current
settings.
1. Select Derivative order: 1
st
, 2
nd
, or 3
rd
derivative.
2. Select Polynomial order: Quadratic or Cubic.
3. Enter the number of points to include in the sub-models in the Points in each
sub-model field. This number is default 15 and has to be odd and >5.
4. Optionally enter the Distance between each point.
5. Click Next.
Note: The filtered dataset is created without the edge effects leaving the first
and last window size/2 variables empty (default first and last 7 variables).

Creating the filtered dataset
The last page completes the creation of the filtered dataset.
- Enter the Dataset name.
- Select to split out unfiltered variables to a new dataset by selecting the Split
out unfiltered variables to a new dataset check box if desired.
- Click Finish.

Note: To create a model using both filtered and original data, both datasets
have to be selected in the Select Data page in the Workset dialog.
User Guide to SIMCA
234
Multiplicative Signal Correction MSC
When applying Multiplicative Signal Correction, each observation (spectra) x
i
is
normalized by regressing it against the average spectrum. For more see the
Multiplicative Signal Correction (MSC) subsection in Appendix B: Preprocessing.
When selecting MSC in the Spectral Filters dialog and clicking Next, the first page of
the MSC wizard opens displaying lists of variables and observations.

Selecting the variables and observations to transform
To apply the filter to the dataset, on the first page of the wizard:
1. In the Variables list, mark and Exclude unwanted variables. All variables
specified as X in the default workset are by default included, while those
specified as Y are by default excluded. Qualitative variables cannot be
filtered.
2. In the Observations list, mark and Exclude the observations to exclude from
the filtering. All observations are by default included.
3. To specify variables and observations according to a fitted model, select the
model in the As model box. Variables in the model specified as Y are
automatically excluded.
4. Click Next.
Creating the filtered dataset
The last page completes the creation of the filtered dataset.
- Enter the Dataset name.
- Select to split out unfiltered variables to a new dataset by selecting the Split
out unfiltered variables to a new dataset check box if desired.
- Click Finish.

Data
235
Note: To create a model using both filtered and original data, both datasets
have to be selected in the Select Data page in the Workset dialog.
Standard Normal Variate SNV
When applying the SNV filter, each observation (spectra) is normalized by
subtracting the mean and dividing with the standard deviation.
For more, see the Standard Normal Variate (SNV) section in Appendix B:
Preprocessing.
Note: In the Workset dialog, select the filtered dataset and then specify the
workset for the calibration model.
SNV wizard
When selecting SNV in the Spectral Filters dialog and clicking Next, the first page of
the SNV wizard opens displaying lists of variables and observations.

Selecting the variables and observations to transform
To apply the filter to the dataset, on the first page of the wizard:
1. In the Variables list, mark and Exclude unwanted variables. All variables
specified as X in the default workset are by default included, while those
specified as Y are by default excluded. Qualitative variables cannot be
filtered.
2. In the Observations list, mark and Exclude the observations to exclude from
the filtering. All observations are by default included.
3. To specify variables and observations according to a fitted model, select the
model in the As model box. Variables in the model specified as Y are
automatically excluded.
4. Click Next.
User Guide to SIMCA
236
Creating the filtered dataset
The last page completes the creation of the filtered dataset.
- Enter the Dataset name.
- Select to split out unfiltered variables to a new dataset by selecting the Split
out unfiltered variables to a new dataset check box if desired.
- Click Finish.

Note: To create a model using both filtered and original data, both datasets
have to be selected in the Select Data page in the Workset dialog.
Row Center
Applying the Row Center filter subtracts the row mean from each row value.
When selecting Row Center in the Spectral Filters dialog and clicking Next, the first
page of the Row Center wizard opens displaying lists of variables and observations.

Selecting the variables and observations to transform
To apply the filter to the dataset, on the first page of the wizard:
1. In the Variables list, mark and Exclude unwanted variables. All variables
specified as X in the default workset are by default included, while those
specified as Y are by default excluded. Qualitative variables cannot be
filtered.
2. In the Observations list, mark and Exclude the observations to exclude from
the filtering. All observations are by default included.
3. To specify variables and observations according to a fitted model, select the
model in the As model box. Variables in the model specified as Y are
automatically excluded. Click Next.
Data
237
Creating the filtered dataset
The last page completes the creation of the filtered dataset.
- Enter the Dataset name.
- Select to split out unfiltered variables to a new dataset by selecting the Split
out unfiltered variables to a new dataset check box if desired.
- Click Finish.

Note: To create a model using both filtered and original data, both datasets
have to be selected in the Select Data page in the Workset dialog.
Savitzky-Golay
Applying the Savitzky-Golay filter removes noise by applying a moving polynomial
to the data. For more, see the Derivatives section in Appendix B: Preprocessing.
When selecting Savitzky-Golay in the Spectral Filters dialog and clicking Next, the
first page of the Savitzky-Golay wizard opens displaying lists of variables and
observations.

User Guide to SIMCA
238
Selecting the variables and observations to transform
To apply the filter to the dataset, on the first page of the wizard:
1. In the Variables list, mark and Exclude unwanted variables. All variables
specified as X in the default workset are by default included, while those
specified as Y are by default excluded. Qualitative variables cannot be
filtered.
2. In the Observations list, mark and Exclude the observations to exclude from
the filtering. All observations are by default included.
3. To specify variables and observations according to a fitted model, select the
model in the As model box. Variables in the model specified as Y are
automatically excluded.
4. Click Next.
Specifying parameters for Savitzky-Golay
Specify the parameters for the filtering in the Filter specification section by following
the steps described here. Note that marking an observation updates the graphs
displaying the Original data (spectra) and the Filtered data with the current settings.
1. Select Polynomial order: Quadratic or Cubic.
2. Enter number of points to include in the sub-models in the Points in each
sub-model field. This number is default 15 and has to be odd and >5.
3. Click Next.
Note: The filtered dataset is created without the edge effects leaving the first
and last window size/2 variables empty (default first and last 7 variables).

Data
239
Creating the filtered dataset
The last page completes the creation of the filtered dataset.
- Enter the Dataset name.
- Select to split out unfiltered variables to a new dataset by selecting the Split
out unfiltered variables to a new dataset check box if desired.
- Click Finish.

Note: To create a model using both filtered and original data, both datasets
have to be selected in the Select Data page in the Workset dialog.
EWMA
Filter removing noise by applying an exponentially weighted moving average to the
data.
When selecting EWMA in the Spectral Filters dialog and clicking Next, the first page
of the EWMA wizard opens displaying lists of variables and observations.

User Guide to SIMCA
240
Selecting the variables and observations to transform
To apply the filter to the dataset, on the first page of the wizard:
1. In the Variables list, mark and Exclude unwanted variables. All variables
specified as X in the default workset are by default included, while those
specified as Y are by default excluded. Qualitative variables cannot be
filtered.
2. In the Observations list, mark and Exclude the observations to exclude from
the filtering. All observations are by default included.
3. To specify variables and observations according to a fitted model, select the
model in the As model box. Variables in the model specified as Y are
automatically excluded.
4. Click Next.
Specifying lambda for EWMA
Enter the desired lambda for the filtering in the Lambda field. Leaving the Lambda
field empty results in using the estimated lambda.
Note that marking an observation updates the graphs displaying the Original data
(spectra) and the Filtered data with the current settings.

Data
241
Creating the filtered dataset
The last page completes the creation of the filtered dataset.
- Enter the Dataset name.
- Select to split out unfiltered variables to a new dataset by selecting the Split
out unfiltered variables to a new dataset check box if desired.
- Click Finish.

Note: To create a model using both filtered and original data, both datasets
have to be selected in the Select Data page in the Workset dialog.
Wavelet Compression Spectral WCS
The wavelet transform of a signal x(t) decomposes the signal x(t) into sets of
coefficients by frequency band. These bands are on a logarithmic scale, and decrease
by the power of 2 from the Nyquist frequency to the lowest frequency in the signal. To
compress the signal, all small coefficients are removed and the transformed signal
compressed. To denoise the signal, all small coefficients are removed and the data is
transformed back to the original domain.
For more, see the Wavelet compression or de-noising of signals section in Appendix
B: Preprocessing.
Note: With compression, reconstruct is done when the Reconstruct wavelets
check box is selected in the Fit page of the Project Options. This is the SIMCA
default.
Digitized spectra
When the dataset consist of digitized spectra (i.e. NIR, NMR, and Raman etc.) each
observation in the X matrix is the spectrum of a sample, and the wavelet transform and
denoising of the matrix X is done row wise.
User Guide to SIMCA
242
WCS wizard
When selecting Wavelet Compression in the Spectral Filters dialog and clicking
Next, the first page of the WCS wizard opens displaying lists of variables and
observations.

Selecting variables, observations, and to transform - WCS
To apply the filter to the dataset, on the first page of the wizard:
1. In the Variables list, mark and Exclude unwanted variables. All variables
are specified according to the default workset.
2. In the Observations list, mark and Exclude the observations to exclude from
the filtering. All observations are by default included.
3. To transform variables prior to the wavelet transformation, select the
Change transformation check box. By default the variables are not
transformed. See the Selecting variable transformations in spectral
filtering subsection next.
4. To specify variables and observations according to a fitted model, select the
model in the As model box. Variables in the model specified as Y are
automatically excluded.
5. Click Next.
Data
243
Selecting variable transformations in spectral filtering
With the Change transformation check box selected, clicking Next opens the Select
transformations page.
Apply transformations before the compression by:
1. Marking the variables to transform.
2. Selecting the transformation: Linear, Log, NegLog, Logit, Exponential, or
Power and clicking the Set-button.
3. Clicking Next.

For details about the transformations available, see the Applying transformations
section in the Workset section of the Home chapter.
Selecting wavelet options
The page for wavelet options holds options for detrending, wavelet function, wavelet
order, exclusion criteria, and compression method. This page is the same for WDS and
WCS.

User Guide to SIMCA
244
In the page for selecting wavelet options:
1. In the Detrend mode box the default is None. Click the box to select Mean
or Linear. Mean signifies removing the mean while Linear signifiers
removing the best linear fit.
2. In the Wavelet function box the default is Daubechies. Click the box to
select Beylkin, Coiflet, Symmlet, Biorthogonal1, Biorthogonal2,
Biorthogonal3, Biorthogonal4, Biorthogonal5, or Biorthogonal6. For
details about the wavelet functions, see the Wavelet families section in
Appendix B: Preprocessing.
3. In the Wavelet order box, select the wavelet order. The default is the lowest
available. The wavelet order differs depending on the wavelet function
selected. Find the orders available for each wavelet function in the wavelet
table.
4. Select the method of exclusion of wavelets coefficients by selecting Energy
retained to be By variance or By detail level.
5. Click the desired decomposition and Compression method, DWT or Best
basis. DWT, discrete wavelet transform, is recommended for low frequency
signals and Best Basis for high frequency signals. With By detail level,
DWT is the only available compression method.
6. Click Next.
Wavelet function Wavelet order
Beylkin N/A
Coiflet 2, 3, 4, 5
Daubechies 4, 6, 8, 10, 12, 20, 50
Symmlet 4, 6, 8, 10
Biorthogonal1 1, 3, 5
Biorthogonal2 2, 4, 6, 8
Biorthogonal3 1, 3, 5, 7, 9
Biorthogonal4 4
Biorthogonal5 5
Biorthogonal6 8

Data
245
Wavelet compression/denoising By variance
When compressing/denoising the signal by the percent of variance explained, SIMCA
computes and displays the percentage of the variance explained by the wavelet
coefficients of the target vector, sorted in order of importance.
1. Use the plot to decide the number of coefficients to keep. The default is to
keep the coefficients that accounts for 99.5% of the variance.
2. Optionally change from the default by clicking another option under Select
by energy retained (sum of squares).
3. Click Next.

Compression/denoising By variance is done either using compression method DWT
or Best Basis. For details about DWT and Best Basis, see the Wavelet compression
or de-noising of signals section in Appendix B: Preprocessing.
Wavelet compression/denoising By detail level
The compression/denoising By detail level is performed by excluding all the detail
coefficients corresponding to selected scales (levels) of the target vector. DWT is the
only available compression method.
The wavelet transform is first done on the target vector and the results are displayed as
a table. After selecting the wavelets coefficients or the scales to be removed, the same
transformation is performed on each X observation vector.
User Guide to SIMCA
246
Selecting wavelet coefficients
When selecting By detail level to compress/denoise the X block by removing detail
levels, SIMCA decomposes the target X to its detail coefficients at every scale. The
left table displays for every detail level (scale) the number of coefficients and the
percentage of the sum of square (energy), not including the DC component.
To exclude levels from the denoised X-block:
1. Select the undesired levels in the Selected list.
2. Click <= to remove.
3. Click Next.

Creating the filtered dataset
The last page completes the creation of the filtered dataset.
- Enter the Dataset name.
- Select to split out unfiltered variables to a new dataset by selecting the Split
out unfiltered variables to a new dataset check box if desired.
- Click Finish.

Note: To create a model using both filtered and original data, both datasets
have to be selected in the Select Data page in the Workset dialog.
WCS dataset properties
When variables have been compressed, the new variables are linear combinations of
the original ones. Loading, coefficients, VIP or any plots displaying variables is
difficult to interpret in the wavelet domain. A property of the wavelet transform is the
fact that it is possible to reconstruct not only the original variables, but also
individual vectors such as loadings, coefficients, VIP etc.
Data
247
By default the Reconstruct wavelets check box is selected (in Project Options), and
all plots, including quick info plots observation wise, are displayed in the domain of
the original dataset.
Use the Workset commands, as usual, to specify the workset for the model.
Wavelet Denoise Spectral WDS
Wavelet Denoise Spectral, WDS, is similar to Wavelet Compression Spectral,
WCS. The difference is that after the removal of the wavelet coefficients, the X block
is transformed back to the original domain for WDS while WCS remains in the
transformed domain but can be reconstructed.
Note: The denoised dataset is transformed back to the original domain.
Therefore reconstruction is not available.

Note: In a WDS transformed dataset, use the Workset commands, as usual, to
specify the workset for the model.
Digitized spectra
When the dataset consist of digitized spectra (i.e. NIR, NMR, and Raman etc.) each
observation in the X matrix is the spectrum of a sample, and the wavelet transform and
denoising of the matrix X is done row wise.
WDS wizard
When selecting Wavelet Denoising in the Spectral Filters dialog and clicking Next,
the first page of the WDS wizard opens displaying lists of variables and observations.

User Guide to SIMCA
248
Selecting the variables and observations to transform
To apply the filter to the dataset, on the first page of the wizard:
1. In the Variables list, mark and Exclude unwanted variables. All variables
specified as X in the default workset are by default included, while those
specified as Y are by default excluded. Qualitative variables cannot be
filtered.
2. In the Observations list, mark and Exclude the observations to exclude from
the filtering. All observations are by default included.
3. To specify variables and observations according to a fitted model, select the
model in the As model box. Variables in the model specified as Y are
automatically excluded. Click Next.
Selecting wavelet options
The page for wavelet options holds options for detrending, wavelet function, wavelet
order, exclusion criteria, and compression method. This page is the same for WDS and
WCS.

In the page for selecting wavelet options:
1. In the Detrend mode box the default is None. Click the box to select Mean
or Linear. Mean signifies removing the mean while Linear signifiers
removing the best linear fit.
2. In the Wavelet function box the default is Daubechies. Click the box to
select Beylkin, Coiflet, Symmlet, Biorthogonal1, Biorthogonal2,
Biorthogonal3, Biorthogonal4, Biorthogonal5, or Biorthogonal6. For
details about the wavelet functions, see the Wavelet families section in
Appendix B: Preprocessing.
3. In the Wavelet order box, select the wavelet order. The default is the lowest
available. The wavelet order differs depending on the wavelet function
selected. Find the orders available for each wavelet function in the wavelet
table.
4. Select the method of exclusion of wavelets coefficients by selecting Energy
retained to be By variance or By detail level.
5. Click the desired decomposition and Compression method, DWT or Best
basis. DWT, discrete wavelet transform, is recommended for low frequency
signals and Best Basis for high frequency signals. With By detail level,
DWT is the only available compression method.
6. Click Next.
Wavelet function Wavelet order
Beylkin N/A
Coiflet 2, 3, 4, 5
Data
249
Wavelet function Wavelet order
Daubechies 4, 6, 8, 10, 12, 20, 50
Symmlet 4, 6, 8, 10
Biorthogonal1 1, 3, 5
Biorthogonal2 2, 4, 6, 8
Biorthogonal3 1, 3, 5, 7, 9
Biorthogonal4 4
Biorthogonal5 5
Biorthogonal6 8
Wavelet compression/denoising By variance
When compressing/denoising the signal by the percent of variance explained, SIMCA
computes and displays the percentage of the variance explained by the wavelet
coefficients of the target vector, sorted in order of importance.
1. Use the plot to decide the number of coefficients to keep. The default is to
keep the coefficients that accounts for 99.5% of the variance.
2. Optionally change from the default by clicking another option under Select
by energy retained (sum of squares).
3. Click Next.

Compression/denoising By variance is done either using compression method DWT
or Best Basis. For details about DWT and Best Basis, see the Wavelet compression
or de-noising of signals section in Appendix B: Preprocessing.
User Guide to SIMCA
250
Wavelet compression/denoising By detail level
The compression/denoising By detail level is performed by excluding all the detail
coefficients corresponding to selected scales (levels) of the target vector. DWT is the
only available compression method.
The wavelet transform is first done on the target vector and the results are displayed as
a table. After selecting the wavelets coefficients or the scales to be removed, the same
transformation is performed on each X observation vector.
Selecting wavelet coefficients
When selecting By detail level to compress/denoise the X block by removing detail
levels, SIMCA decomposes the target X to its detail coefficients at every scale. The
left table displays for every detail level (scale) the number of coefficients and the
percentage of the sum of square (energy), not including the DC component.
To exclude levels from the denoised X-block:
1. Select the undesired levels in the Selected list.
2. Click <= to remove.
3. Click Next.

Creating the filtered dataset
The last page completes the creation of the filtered dataset.
- Enter the Dataset name.
- Select to split out unfiltered variables to a new dataset by selecting the Split
out unfiltered variables to a new dataset check box if desired.
- Click Finish.

Note: To create a model using both filtered and original data, both datasets
have to be selected in the Select Data page in the Workset dialog.
Data
251
Orthogonal Signal Correction OSC
OSC uses a PLS model, to remove from X, information that is orthogonal to Y. Results
are displayed as ordinary PLS plots, i.e. scores, loading. These plots are interpreted as
regular PLS plots, with the difference that the patterns displayed are orthogonal to,
rather than correlated to, Y.
Note: The OSC filter is prone to overfit and may give results that are too
optimistic. The OPLS/O2PLS approach provides more realistic results and
separates orthogonal variation from predictive variation in a single model.
When selecting OSC in the Spectral Filters dialog and clicking Next, the first page of
the OSC wizard opens displaying lists of variables and observations.
Note: Serious outliers in the score plot need to be removed before OSC, as
they may bias the corrected data X.

Selecting variables, observations, to transform, and to scale
- OSC
To apply the filter to the dataset, on the first page of the wizard:
1. In the Variables list, specify the Y variables and mark and Exclude
unwanted variables. All variables are specified according to the default
workset.
2. In the Observations list, mark and Exclude the observations to exclude from
the filtering. All observations are by default included.
3. To transform variables prior to the OSC, select the Change transformation
check box. For more, see the Selecting variable transformations in
spectral filtering subsection next.
User Guide to SIMCA
252
4. To change the scaling of the variables, select the Change scaling check box.
By default, all the X variables are centered and the Y variables are centered
and scaled to unit variance UV. For more, see the Selecting scaling in
spectral filtering subsection later in this chapter.
5. To specify variables and observations according to a fitted model, select the
model in the As model box.
6. Click Next.
Selecting variable transformations in spectral filtering
With the Change transformation check box selected, clicking Next opens the Select
transformations page.
Apply transformations before the compression by:
1. Marking the variables to transform.
2. Selecting the transformation: Linear, Log, NegLog, Logit, Exponential, or
Power and clicking the Set-button.
3. Clicking Next.

For details about the transformations available, see the Applying transformations
section in the Workset section of the Home chapter.
Selecting scaling in spectral filtering
With the Change scaling check box selected, clicking Next opens the Select scaling
page.
To change scaling:
1. Mark the variables.
2. Click the desired scaling button: None, Pareto, Unit Variance, or Center.
3. Click Next.
Data
253

Computing the OSC model
SIMCA computes the first component of the OSC model, and displays the angle
between t and Y for that component, the Sum of Squares (SS) remaining in the X
block, and the Eigenvalue.
When the angle is 90 degree, OSC has reached orthogonality.
- To compute the next component, click Next Component. Usually two
components are recommended.
- When done, select the number of components to retain in the Components
to use field.
- Click Next.

Creating the filtered dataset
The last page completes the creation of the filtered dataset.
- Enter the Dataset name.
- Select to split out unfiltered variables to a new dataset by selecting the Split
out unfiltered variables to a new dataset check box if desired.
- Click Finish.

User Guide to SIMCA
254
Note: To create a model using both filtered and original data, both datasets
have to be selected in the Select Data page in the Workset dialog.
OSC dataset
With an OSC dataset, use the menu Plot/List to examine the OSC model by:
1. Selecting a plot type.
2. Selecting the OSC-dataset in the Data box.
3. Selecting Variable and Scores or Observations and Loadings in the Select
data type box.
4. Adding the series to plot.


Vector Description
OSC-p "Loadings of the X variables in the OSC model.
These best approximate X together with scores that are
orthogonal to Y."
OSC-t X scores t of the OSC model. They summarize X and are
the best combination of the X's that are orthogonal to Y.
OSC-w Loading weights of the X variables or their residuals (higher
dimensions) in the OSC model. These weights are selected
to make t orthogonal to u.
OSC-w* Loading weights of the X variables (not their residuals) in
the OSC model. X variables with large w* are little
correlated with u and (Y).
Fitting a calibration model
To fit a calibration model, create a new workset and select both the OSC dataset and
the original dataset, or the split dataset, to access the y-variables. Any scaling or
transformation specified during the filtering needs to be specified in the workset.
To fit the PLS calibration model to the OSC data, click Autofit.
Data
255
Using OSC with a batch level dataset
When you have quality variables in a batch level dataset, and want to use a PLS model,
you may want to use OSC to preprocess your data. After using OSC, SIMCA creates a
new OSCed batch level dataset.
To OSC preprocess the predictionset, select the batch evolution predictionset.
Preprocessing is then done automatically.
Chaining filters
Apply several filters in combination by clicking Spectral Filters on the Data tab and
adding the desired filters. The filters can be applied in any order although each filter
can be applied only once.
Note: If wavelet compression is performed before OSC filtering, the former
usually needs to keep 99.9 of the variance.

To define the chain of filters:
1. Click a filter in the Available list.
2. Click the => button.
3. Repeat 1-2 until the desired chain of filters is displayed in the Selected list.
4. Click OK and the wizard of each filter opens successively and the filters are
applied in the order specified in the Spectral Filters dialog.
Predictionsets and filtered datasets
When the model is built on a filtered dataset, the predictionset may not be filtered. The
predictionset is automatically signal corrected/denoised as the filtered dataset.
Use the Predict tab commands, as usual.
Note: The predictionset needs to include all variables of the original dataset
the filtered dataset was created from.
User Guide to SIMCA
256
Time Series Filters
The following time series filters are available by clicking Time Series Filters on the
Data tab: Wavelet Compress Time Series (WCTS) and Wavelet
Denoising/Decimation (WDTS). These filters are applied variable wise.

A short background is available in PLS wavelet compression of time series section in
Appendix B: Preprocessing.
Wavelet Compress Time Series WCTS
With continuous process data (time series), you may want to compress the data to
reduce the number of observations.
The wavelet transform of time series focuses on the selected target response Y, with
the objective to achieve a parsimonious representation of the signal Y, while keeping
all the information.
Note: The dataset must hold at least one y-variable to wavelet compress time
series, as the compression is designed for PLS models.
The Wavelet Compress Time Series differs from Wavelet Compress Spectral in that
compression is done variable wise and not observation wise and that WCTS requires a
y-variable.
Note: With compression, reconstruct is done when the Reconstruct wavelets
check box is selected in the Fit page of the File | Project Options. This is the
SIMCA default.
Compression method
The X and Y blocks are first mean centered and scaled to UV and then transformed,
using DWT or Best basis, to the wavelet domain. After compression the mean of
every variable and original variance are added back to the wavelet coefficients of that
variable.
WCTS wizard
To compress the data by reducing the number of observations, in the Dataset tab, click
Time Series Filters | Wavelet Compress Time Series. The first page of the WCTS
wizard opens displaying lists of variables and observations.
Data
257

Selecting y-variables and observations - WCTS
To apply the filter to the dataset, on the first page of the wizard:
1. In the Variables list, specify the Y variables and mark and Exclude
unwanted variables. With many y-variables they must be positively
correlated, and the most important one should be selected as target on the
next page. All variables are specified according to the default workset.
2. In the Observations list, mark and Exclude the observations to exclude from
the filtering. All observations are by default included.
3. To specify variables and observations according to a fitted model, select the
model in the As model box.
4. Click Next.
Note: Specifying several uncorrelated y-variables or anti-correlated y-
variables as Y will give poor results. In such cases first do a PCA and select
only a group of positively correlated y-variables to be compressed together.
Selecting wavelet options
The page for wavelet options holds options for detrending, wavelet function, wavelet
order, exclusion criteria, compression method, and target variable. This page is the
same for WCTS and WDTS.
User Guide to SIMCA
258

In the page for selecting wavelet options:
1. In the Detrend mode box the default is None for WDTS. Click the Detrend
mode box to select Mean or Linear. Mean signifies removing the mean
while Linear signifiers removing the best linear fit. For WCTS the Detrend
mode box is unavailable and the selected detrending mode is Mean.
2. In the Wavelet function box the default is Daubechies. Click the Wavelet
function box to select Beylkin, Coiflet, Symmlet, Biorthogonal1,
Biorthogonal2, Biorthogonal3, Biorthogonal4, Biorthogonal5, or
Biorthogonal6. For details about the wavelet functions, see the Wavelet
families section in Appendix B: Preprocessing.
3. In the Wavelet order box, select the wavelet order. The default is the lowest
available. The wavelet order differs depending on the wavelet function
selected. Find the orders available for each wavelet function in the table in
the Selecting wavelet options - WCS subsection earlier in this chapter.
4. Select the method of exclusion of wavelets coefficients by selecting Energy
retained type By variance or By detail level.
5. Click the desired decomposition and Compression method, DWT or Best
basis. DWT, discrete wavelet transform, is recommended for low frequency
signals and Best Basis for high frequency signals. With By detail level,
DWT is the only available compression method.
6. In the Select target variable box select which of the variables to use as
target variable.
7. Click Next.
Wavelet compression By variance - WCTS
When compressing the signal by the percent of variance explained, SIMCA computes
and displays the percentage of the variance explained by the wavelet coefficients of the
target vector, sorted in order of importance.
1. Use the plot to decide the number of coefficients to keep. The default is to
keep the coefficients that account for 90% of the variance. The same
coefficients are kept for all y-variables.
2. Optionally change from the default under Select by energy retained (sum of
squares).
3. Click Next.
Data
259

Compression/denoising By variance is done either using compression method DWT
or Best Basis. For details about DWT and Best Basis, see the Wavelet transform
section in Appendix B: Preprocessing.
Wavelet compression/denoising By detail level
The compression/denoising By detail level is performed by excluding all the detail
coefficients corresponding to selected scales (levels) of the selected target vector.
DWT is the only available compression method.
The wavelet transform is first done on the target vector (average variable) and the
results are displayed as a table. After selecting the wavelets coefficients or the scales to
be removed, the same transformation is performed on each X observation vector.
Selecting wavelet coefficients
When selecting By detail level to compress/denoise the X block by removing detail
levels, SIMCA decomposes the target X to its detail coefficients at every scale. The
left table displays for every detail level (scale) the number of coefficients and the
percentage of the sum of square (energy), not including the DC component.
To exclude levels from the denoised X-block:
1. Select the undesired levels in the Selected list.
2. Click <= to remove.
3. Click Next.
User Guide to SIMCA
260

Creating the new dataset
On the last page of the WCTS wizard:
- Enter the Dataset name.
- Click Finish.

WCTS dataset properties
The X and Y matrices are compressed by extracting only the selected number of
significant coefficients, or detail level coefficients, in both the X and Y block, column
wise.
Observation identifiers with Best Basis
With Best Basis selected, the observations in the new project are labeled $ WTS-
ScalePosition (e.g., $WTS-6-15).
Scale is the block index of the packet table. It represents the frequency resolution of
the signal, i.e., the number of times it has passed through the Low pass and Hi pass
filters. This means that when the scale value equals 1 the signal has passed once
through the filters. The scale ranges from 1 to log2 (N), where N=length of the signal
(padded if N was not power of 2). Frequency resolution and time resolutions are the
inverse of each other.
Position is the position of the coefficient within the block. The position index ranges
from 1 to N, and is related to the frequencies in the signal. Here 1 is the lowest and N
is the highest frequency.
Predictionset
Predictionsets will not be wavelet transformed, and can be used as they are in the
original domain.
Data
261
Wavelet Denoising/Decimation WDTS
The Wavelet Denoising/Decimation differs from Wavelet Denoise Spectral in that
compression is done variable wise and not observation wise.
After the removal of the wavelet coefficients, the X variables are transformed back to
the original domain.
After transforming all the variables to the wavelet domain, the denoising is done by
performing the inverse wavelet transform with either the selected coefficients or the
coefficients from the selected details. All other coefficients are set to 0.
Decimation
Decimation is a re-sampling of data after removal of the high frequencies, by selecting
every second or fourth or eighth, etc. observations. This is useful when the sampling
interval is much faster than the time constant of the system.
The decimation is performed on the dataset after it is transformed back to original
units. Decimation is only available by a value that is power of 2.
WDTS wizard
To compress the data by reducing the number of observations, on the Data tab, in the
Filters group, click Time Series Filters | Wavelet Denoising/Decimation. The first
page of the WDTS wizard opens displaying lists of variables and observations.

Selecting the variables and observations to transform
To apply the filter to the dataset, on the first page of the wizard:
User Guide to SIMCA
262
1. In the Variables list, mark and Exclude unwanted variables. All variables
specified as X in the default workset are by default included, while those
specified as Y are by default excluded. Qualitative variables cannot be
filtered.
2. In the Observations list, mark and Exclude the observations to exclude from
the filtering. All observations are by default included.
3. To specify variables and observations according to a fitted model, select the
model in the As model box. Variables in the model specified as Y are
automatically excluded.
4. Click Next.
Selecting wavelet options
The page for wavelet options holds options for detrending, wavelet function, wavelet
order, exclusion criteria, compression method, and target variable. This page is the
same for WCTS and WDTS.

In the page for selecting wavelet options:
1. In the Detrend mode box the default is None for WDTS. Click the Detrend
mode box to select Mean or Linear. Mean signifies removing the mean
while Linear signifiers removing the best linear fit. For WCTS the Detrend
mode box is unavailable and the selected detrending mode is Mean.
2. In the Wavelet function box the default is Daubechies. Click the Wavelet
function box to select Beylkin, Coiflet, Symmlet, Biorthogonal1,
Biorthogonal2, Biorthogonal3, Biorthogonal4, Biorthogonal5, or
Biorthogonal6. For details about the wavelet functions, see the Wavelet
families section in Appendix B: Preprocessing.
3. In the Wavelet order box, select the wavelet order. The default is the lowest
available. The wavelet order differs depending on the wavelet function
selected. Find the orders available for each wavelet function in the table in
the Selecting wavelet options - WCS subsection earlier in this chapter.
4. Select the method of exclusion of wavelets coefficients by selecting Energy
retained type By variance or By detail level.
5. Click the desired decomposition and Compression method, DWT or Best
basis. DWT, discrete wavelet transform, is recommended for low frequency
signals and Best Basis for high frequency signals. With By detail level,
DWT is the only available compression method.
6. In the Select target variable box select which of the variables to use as
target variable.
7. Click Next.
Data
263
Wavelet denoising By variance - WDTS
When denoising the signal by the percent of variance explained, SIMCA computes and
displays the percentage of the variance explained by the wavelet coefficients of the
target vector, sorted in order of importance.
1. Use the plot to decide the number of coefficients to keep. The default is to
keep the coefficients that accounts for 99.5% of the variance.
2. Optionally change from the default under Select by energy retained (sum of
squares).
3. Click Next.

Compression/denoising By variance is done either using compression method DWT
or Best Basis. For details about DWT and Best Basis, see the Wavelet transform
section in Appendix B: Preprocessing.
Wavelet compression/denoising By detail level
The compression/denoising By detail level is performed by excluding all the detail
coefficients corresponding to selected scales (levels) of the selected target vector.
DWT is the only available compression method.
The wavelet transform is first done on the target vector (average variable) and the
results are displayed as a table. After selecting the wavelets coefficients or the scales to
be removed, the same transformation is performed on each X observation vector.
User Guide to SIMCA
264
Selecting wavelet coefficients
When selecting By detail level to compress/denoise the X block by removing detail
levels, SIMCA decomposes the target X to its detail coefficients at every scale. The
left table displays for every detail level (scale) the number of coefficients and the
percentage of the sum of square (energy), not including the DC component.
To exclude levels from the denoised X-block:
1. Select the undesired levels in the Selected list.
2. Click <= to remove.
3. Click Next.

Creating the new dataset - WDTS
On the last page of the WDTS wizard:
- Enter the Dataset name.
- Optionally enter the number to decimate. This option is available when By
detail level was selected in the Selecting wavelet options page. For more,
see the Applying decimation subsection next.
- Click Finish.
Data
265

Applying decimation
Decimation is available when using By detail level. The objective is to select a subset
of the observations by keeping every second or fourth or eighth etc. observation.
Decimation is only available by a value that is power of 2. For example select 4, every
4th observation will be used. Leaving the decimation = 1 will include all observations.
The following decimation is recommended:
- After removing D1, decimate by 2.
- After removing D1 and D2, decimate by 4
- After removing D1, D2, and D3, decimate by 8 etc.
Note: Decimation is done after the WDTS transformation.
WDTS dataset properties
The dataset is compressed by extracting only the selected number of significant
coefficients, or detail level coefficients, column wise.
In the WDTS transformed project, use the Workset commands, as usual, to specify the
workset for the model.
Predictionset
Predictionsets will not be wavelet transformed, and can be used as they are in the
original domain.
User Guide to SIMCA
266
Dataset Summary
The Dataset Summary, or Properties dialog, holds information about the selected
dataset including number of variables and observations, where the data was imported
from, missing values, variable types etc.
To open the dialog:
- On the Data tab, in the Summary group, click Dataset Summary.
- Right-click the open dataset spreadsheet and click Properties.
In the dataset properties dialog there are always three tabs Variables, Observations
and General.
For batch evolution datasets the following tabs are additionally available:
- Batches tab always.
- Phases tab when there are phases.
For filtered datasets the Filter tab is additionally available.

Data
267
General page in dataset Properties
The General page displays the number of variables and observations, number of
batches and phases for batch projects, and the Import Log.

Observations page
To display the observations and assigning them to classes for the default workset, click
the Observations tab. After assigning classes in here, the classes are by default
assigned in all new worksets.
To view secondary IDs, right-click the list, and select the desired ID.
To assign observations to a class:
1. Mark the observations manually or using Find.
2. Type the class name in the Class field or use class numbers.
3. Click Set.
4. Repeat 1-3 until done.
5. Click Apply or OK.

Note: Assigning classes is unavailable in batch projects.
User Guide to SIMCA
268
Variables page
To display the variables and their roles (i.e. what's X and what's Y), click the Variable
tab.
To change the variable type, open the Workset dialog, modify as desired, and then
click the Save as Default Workset button.

Date/Time configuration in Dataset Properties and Dataset
Summary
A variable that has been specified as Y Date/Time Variable or Date/Time Variable
can be reformatted by right-clicking the variable in the dataset spreadsheet, and then
clicking Format Date/Time. See step 4 in the table.
Reformatting can also be done, as well as other Date/Time variable settings, in
Dataset Properties dialog described in the table, opened by clicking Dataset
Summary in the Summary group on the Data tab.
In the Date/Time configuration section there are three buttons. The No Default
button is always available while the Edit Format (described in the table) and Set As
Default, buttons are available when a date/time formatted variable is selected in the
list.
Clicking the Set As Default-button sets the selected Date/Time variable as the default
variable to display in line plots and scatter plots with one series.
Clicking the No Default-button will revert to using Num in the places where
Date/Time is by default displayed when available.
Note: Editing the format only affects how the variable is displayed, not how it
is saved.
Data
269

Step Dialog
1. Open the Dataset
Properties dialog and click
the Variables tab.
2. Mark the date/time variable.
3. Click the Edit Format-
button in the Date Time
configuration section.

4. In the Specify Date
Format dialog, enter the
new format.


Phases
The Phases page displays all phases in the dataset with the number of observations and
batches in that phase. Under Comments, view the current y-variable for each phase.
Clicking the + sign displays the variables included in the respective phases.

User Guide to SIMCA
270
Batches
The Batches page displays all batches with the number of observations included in
each batch.

When there are phases, clicking the + sign displays the phases included in the
respective batch.
Clicking the Batch - Phase or Phase - Batch buttons switches what is displayed
between:
- Batches with the phases displayed when clicking the + sign when the mode
is Batch-Phase.
- Phases with the batches displayed when clicking the + sign when the mode is
Phase-Batch.
Filter summary
The Filter page holds details about the performed filtering in the selected preprocessed
dataset. A preprocessed dataset is a dataset created by spectral filtering (Data |
Spectral Filters) or time series filtering (Data | Time Series Filters). The Filter page
is available in the dataset Properties dialog.
To display the Filter page click:
- On the Data tab, in the Summary group, click Dataset Summary and in the
Properties dialog click the Filter tab.
- Right-click the filtered dataset spreadsheet and select Properties.
Data
271
In the example here, the current dataset is the result of having filtered and decimated a
time series dataset, and the Filter page displays the following:

Trimming Overview
The Trimming Overview is available after trimming or Winsorizing the dataset.
To get an overview of the trimming performed, open the Trimming overview.
The Trimming overview is available by:
- On the Data tab, in the Summary group, clicking Trimming Overview.
- Right-clicking the dataset spreadsheet and selecting Trimming Overview.
The Trimming Overview displays a summary of the trimmed or replaced values.
Trimmed values replaced by missing appear in magenta (pink), Winsorized values
appear in green.
Note: The trimming-Winsorizing effects only the selected observations and
variables.

User Guide to SIMCA
272
Displaying the Missing Value Map
To display an overview of a dataset with respect to missing values:
- On the Data tab, in the Summary group, click Missing Value Map and then
the desired dataset in SIMCA.
- In SIMCA import, on the View tab in the Missing Values group, click the
Missing Value Map.
Missing values are colored while data present are white.

Spectra
With spectral data, it is particularly useful to display XObs of all observations in a line
plot, the Spectra plot.
Open the Spectra plot on the Data tab, in the Summary group.

See also the Color subsection in the Tools tab section in the Plot and list contextual
tabs chapter for details about coloring the line plot.
Data
273
Hierarchical models
After fitting a model, the scores, residuals, and predicted y of that model can be used as
variables in another model.
Models using scores, residuals, or predicted y from another model are named
hierarchical top models. The model providing the score/residual/predicted y variables
is named hierarchical base model.
To create the hierarchical variables to use in another model, use one of the two
methods:
- With the model marked in the Project Window, on the Data tab, in the Base
Model section click Hierarchical Base Model and select check boxes as
desired.
- Right-click the model in the Project Window, click Hierarchical Base
Model and select the check boxes as desired.
A new dataset named $MxHierarchical is created holding:
- The scores of all extracted components, when selecting Scores.
- The residuals of all the variables, when selecting the residual check boxes.
- The predictions for all y-variables when selecting Y Pred.
Models including the hierarchical dataset are marked T (hierarchical Top model), and
the Base models are marked B in the Project Window.

User Guide to SIMCA
274
Hierarchical models with OPLS or O2PLS
When the fitted model is an OPLS or O2PLS model, the following additional
hierarchical terms can be added:
- Y-Orthogonal residuals - the residuals after subtracting the predictive part
of the model (X-TP'). Only available for OPLS and O2PLS models with at
least one Y-Orthogonal component.
- Y-Related residuals - the residuals after subtracting the Y-orthogonal part
of the model (X-ToPo'). Only available for OPLS and O2PLS models with at
least one Y-Orthogonal component.
- Y-Orthogonal scores - the scores that are orthogonal to Y (To). Only
available for OPLS and O2PLS models with at least one Y-Orthogonal
component.

Hierarchical models with blocks
After assigning variables to blocks (see Assigning variables to blocks subsection in
the Workset chapter), the hierarchical model types PCA-Hierarchical, and if y-
variables were specified, PLS-Hierarchical, OPLS-Hierarchical and O2PLS-
Hierarchical are available.
Selecting one of the hierarchical model types results in:
- Fitting one model for each block positioned in a CM wrapper.
- Setting each model as hierarchical base model creating the hierarchical
scores datasets.
Hierarchical base models reverts to non hierarchical
A hierarchical base model reverts to a regular Non Hierarchical when:
- Adding components
- Changing the model title
- Clicking Non Hierarchical
This results in that the B is removed and the hierarchical dataset is deleted.
Consequently the hierarchical top models are deleted and the former base model has to
be designated as a Hierarchical Base Model again to recreate the hierarchical top
models.

275
Batch
Introduction
The Batch tab holds the batch specific features and is unavailable for regular projects.
The following features are available:
- Batch control charts (BCC) plots and lists, both for the model and for
predictions.
- Batch plots illustrating smoothing and alignment.
- Creation of batch level and hierarchical batch datasets.
- Batch variable importance plot.

The Batch modeling in SIMCA section, in the Introduction to multivariate data
analysis chapter, describes the batch evolution and batch level model types.
Batch Control Charts
For batch evolution models, batch control charts are available on the Batch tab. The
available batch control chart types are:
- Scores
- Variable
- DModX
- Hotellings T2
- Observed vs. Time/Maturity
- List
All plots are available both in the Analysis Control Charts and the Prediction
Control Charts groups, the buttons in the latter with the addition of PS in the button
name.
Each plot type is described later in this section.
The batch control charts by default display the first batch in the workset. For batch
projects with phases the active phase (class) is used. When the BEM is marked when
clicking a BCC all phases are displayed.
User Guide to SIMCA
276
Note: For any batch control chart with a batch outside the displayed control
limits, you can display the Out of Control plot by clicking Out of Control
Summary in the Change Type group on the Tools tab. For more, see the Out
of Control plot (OOC) subsection in the Plot and list contextual tabs chapter.
Batch control chart background
When batches have different length, alignment of the batch variables, scores, DModX,
etc., is done by cutting or continuing with the last value so that all batches have the
same length as the median batch length. This subsection describes the batch control
chart calculations using, as example, the scores.
Note: With batches that vary in length more than 20% to 30%, you should use
a maturity variable as Y.
The scores of the active model(s) are chopped up, aligned and reorganized so that the
scores of one batch form one row vector (t1 followed by t2, followed by t3, etc.) in a
matrix S
T
. This matrix has N rows (one per batch) and x J (AJ) columns from the A
score vectors and the J "time points" per batch.
When batches have phases, the alignment is done by phases using their respective
Maturity or Time variable.
Time Normalized is equivalent to a linear time warping.
From the matrix S
T
, SIMCA calculates the averages and standard deviations (SD) of
the scores, (average trace of normal batches and control intervals as the averages 2,
and 3 SD).
The batch control charts for Variables, DModX, Hotelling's T2, and Obs vs.
Time/Maturity are calculated in the same manner.
Batch control chart plot and list customization
Batch control charts can be customized using the Properties and Format Plot dialogs.
This section describes the Properties dialog for batch control charts. For more about
Format Plot, see the Plot and list contextual tabs chapter.
Open the Properties dialog by clicking the plot and clicking Properties in the mini-
toolbar
In the Properties dialog for the BCCs the following tabs are available:
- Select Batch and Phase
- Limits and Averages
- Component.
- Select Variable for the Variable BCC.
Note: The Select Batch and Phase page is available for the Batch Control
Chart List.
Each tab is described in the subsections that follow.
Batch
277
Selecting Batch and Phase
On the Select Batch and Phase page:
- In the Select phase box select one phase or All phases.
- In the Unselected list, select the batches to display and click the => button.
- In the Selected list, mark available batches to no longer display and click the
<= button.
- Select or clear the Align batches check box (by default both workset and
predictionset batches are displayed unaligned) to display the batches aligned
or unaligned. After clearing, select what to display on the X-Axis: YVar,
YVarDS, or Num. Or for the batch control charts created from the
Predictions menu: YVarPS, YVarDSPS, or Num. YVar and YVarPS are
the time/maturity variable treated as in the workset; YVarDS and
YVarDSPS are the time/maturity variable in original units (as in the
respective datasets).
Note: Batch Control Charts displaying All Phases use Num on the x-axis.

Limits and Averages
On the Limits and Averages page:
- In the Control limits box, select which limits to display. Default is to display
Average Batch and when appropriate +3 std. dev and -3 std. dev. Available
limits are by default -3, -2, -1, Average Batch, 1, 2, 3 standard deviations.
- To display the standard deviation over all batches and phases, select the
Display steady state limits using the overall standard deviation check
box. This option is available for the Variable BCCs displaying one phase,
and enables monitoring of steady state variables.
- In the For this plot box in the Averages section, select
a. do not remove the average to display the limits and values on
the original scale.
b. remove the average to remove the average from limits and
batches leading to a centered plot with y=0 as average batch.
User Guide to SIMCA
278
c. remove the average and normalize the values to remove the
average and normalize limits and batches leading to a plot with
y=0 as average batch and horizontal limits.

Adding limits
To add control limits, click the arrow next to the Control limits box and click Custom
Limits.


In the Custom Limits dialog, enter the limits to display in Control Limits, with
<space> parting the limits.
Note: The new entered limits replace the old. This means that typing 1.5 in
Limits results in that and +1.5 std. dev will be the only available limits.

To restore the default limits, click the arrow next to the Control limits box and click
Reset Custom Limits to Default.
Batch
279
Component
On the Component page, select for which component to display the batch control
chart. The Component page is not available for the Variable BCC.

Select Variable
On the Select Variable page, select which variable to display in the Variable BCC.

Excluding batches
An entire batch can be excluded from the model by selecting the batch in a plot using
the Batch Marking Mode.
To exclude a batch from the model:
1. Open a plot displaying batches, such as a batch control chart.
2. On the Tools tab, in the Highlight group, click Select | Batch Marking
Mode.
3. In the plot click the batch.
4. On the Marked Items tab, in the Modify Model group, click Exclude.
5. New unfitted models are created.
For details on the tools, see the Select plot items - Marking tool section in the
Marked Items tab section in the Plot and list contextual tabs chapter.
See also the Including and excluding batches subsection in the Batch page section in
the Home chapter.
User Guide to SIMCA
280
Score Batch Control Chart
Open the Scores BCC by clicking it in the Analysis Control Chart group on the
Batch tab. This plot displays the workset batches.
To see predictions, click Scores PS BCC in the Prediction Control Chart group.
The limits that are displayed by default are the Average Batch and -3 and +3 std. dev.

Variable Batch Control Chart
Open the Variable BCC by clicking it in the Analysis Control Chart group on the
Batch tab. This plot displays the workset batches.
To see predictions, click DModX PS BCC in the Prediction Control Chart group.
The limits that are displayed by default are the Average Batch and -3 and +3 standard
deviations.

By selecting the Display steady state limits using the overall standard deviation
check box, in the Properties dialog, Limits and Averages page, you can select to
have the limits computed from the standard deviation of the variable as computed or
specified in the workset.
Batch
281
DModX Batch Control Chart
Open the DModX BCC by clicking it in the Analysis Control Chart group on the
Batch tab. This plot displays the workset batches.
To see predictions, click DModX PS BCC in the Prediction Control Chart group.
The limits that are displayed by default are the Average Batch and +3 std. dev.

Hotellings T2Range Batch Control Chart
Open the Hotellings T2 BCC by clicking it in the Analysis Control Chart group on
the Batch tab. This plot displays the workset batches.
To see predictions, click Hotelling's T2PS BCC in the Prediction Control Chart
group.
The limits that are displayed by default are the Average Batch and T2Crit(95%) and
T2Crit(99%).

User Guide to SIMCA
282
Observed vs. Time/Maturity Batch Control Chart
Open the Observed vs. Time/Maturity BCC by clicking Obs vs. Time/Maturity BCC
in the Analysis Control Chart group on the Batch tab. This plot displays the workset
batches.
To see predictions, click Obs vs. Time/Maturity PS BCC in the Prediction Control
Chart group.
The limits that are displayed by default are the Average Batch and -3 and +3 std. dev.

Batch Control Charts List
The List BCC displays a number of vectors for the active model or phase.
Open the BCC List by clicking it in the Analysis Control Chart group on the Batch
tab. This plot displays the workset batches.
To see predictions, click List PS BCC in the Prediction Control Chart group.

Default is to display the list for unaligned batches. All t, YPred, and DModX vectors
are then displayed unaligned while the averages and standard deviations are displayed
aligned.
Note: The t, YPred, and DModX vectors referring to the specific batches are
displayed alongside the YVarDS vector to the far right in the list.
To display the list for aligned batches, open Properties, and select the Align batches
check box.
Batch
283
All vectors available from the BCC List are listed in the table.
General name Description Example
Mx.t[comp]
(Aligned) (Avg)
The average over all
batches of the aligned score
vectors, for each
component.
M3.t[1] (Aligned) (Avg)
Mx.t[comp]
(Aligned) (Std.
Dev.)
The standard deviation over
all batches of the aligned
score vectors, for each
component.
M3.t[1] (Aligned) (Std. Dev.)
Mx.t[comp]
(Aligned): batch
The aligned score vector for
the listed batch, for each
component.
Note: This vector is only
available when the Align
batches check box is
selected.
M3.t[1] (Aligned): 1
Mx.YPred[last
comp](y)(Aligned)
(Avg)
The average over all
batches of the aligned
predicted y vectors, for the
last component.
M3.YPred[2](timeb)(Aligned)
(Avg)
Mx.YPred[last
comp](y)(Aligned)
(Std. Dev.)
The standard deviation over
all batches of the aligned
predicted y vectors, for the
last component.
M3.YPred[2](timeb)(Aligned)
(Std. Dev.)
Mx.YPred[last
comp](y)(Aligned):
batch
The aligned predicted y
vector for the listed batch
and last component.
Note: This vector is only
available when the Align
batches check box is
selected.
M3.YPred[2](timeb)(Aligned):
1
Mx.DModX[last
comp] (Aligned)
(Avg)
The average over all
batches of the aligned
DModX vectors, for the last
component.
M3.DModX[2] (Aligned)
(Avg)
Mx.DModX[last
comp] (Aligned)
(Std. Dev.)
The standard deviation over
all batches of the aligned
DModX vectors, for the last
component.
M3.DModX[2] (Aligned) (Std.
Dev.)
Mx.DModX[last
comp] (Aligned):
batch
The aligned DModX vector,
for the listed batch and last
component.
Note: This vector is only
available when the Align
batches check box is
selected.
M3.DModX[2] (Aligned): 1
User Guide to SIMCA
284
General name Description Example
Mx.YVarDS(y):
batch
The original y-vector for the
listed batch.
Note: This vector is only
available when the Align
batches check box has
been cleared.
M3.YVarDS(timeb): 1
Mx.t[comp]: batch The unaligned score vector
for the listed batch, for each
component.
Note: This vector is only
available when the Align
batches check box has
been cleared.
M3.t[1]: 1
Mx.YPred[last
comp] (y): batch
The unaligned predicted y
vector for the listed batch
and last component.
Note: This vector is only
available when the Align
batches check box has
been cleared.
M3.YPred[2] (timeb): 1
Mx.DModX[last
comp]: batch
The unaligned DModX
vector for the listed batch
and last component.
Note: This vector is only
available when the Align
batches check box has
been cleared.
M3.DModX[2]: 1
Batch control charts for new batches
To monitor the evolution of new batches, use the buttons in the Prediction Control
Chart group on the Batch tab, after specifying the predictionset in the Specify
Predictionset group on the Predict tab.
The data from each new batch are inserted into the batch evolution model, giving
predicted values of the scores, TPS. In addition, predicted Y and DModX values are
computed. These results can now be plotted in the appropriate control charts, with
limits derived from the workset model.
The batch control charts indicate whether the batch is starting and running normally or
not. If the values are outside the normal ranges, contribution plots based on the x-
values or the residuals indicate which variables together are related to the deviations.
The new batches are by default displayed unaligned in the predicted batch control
charts.
Use the Data group, in the Tools tab, to select a different phase, different batches, etc.
Batch
285

For details about creating contribution plots for entire batches, see the second table in
the Drill down contribution plots available subsection in the Marked Items tab
section in the Plot and list contextual tabs chapter.
For details about the Properties dialog, see the Batch Control Chart plot and list
customization subsection previously in this chapter.
For details about the limits, see the Calculating limits to build control charts
subsection in the Introduction to MVA chapter.
For info about the out of control plot, see the Out of Control Summary Plot (OOC)
subsection in the Tools tab section in the Plot and list contextual tabs chapter.
Create BLM
To be able to create a batch level model, BLM, you first have to create a batch level
dataset, BL DS. The BL DS then resides in the same project as the batch evolution
dataset.
Batch level models, BLM, created using the BL DS, are positioned in the same BM as
the BEM the BL DS was created from.

User Guide to SIMCA
286
Create the BL DS by clicking Create Batch Level in the Dataset group on the Batch
tab. The Create Batch Level Dataset wizard opens.

Selecting what to base the batch level dataset on
On the first page of the Create Batch Level Dataset wizard, you must select what to
base the batch level dataset, BL DS, on. The BL DS can be created using the options
described in the table. Note that all check boxes can be selected and multiple BL DS
are then created simultaneously.
Options Action Result after
Next/Finish
1. Create BL DS from
scores.
Select the Scores from
the active model or
group of models check
box.
A dataset containing
scores is created after
Finish.
The variables are aligned
using the time/maturity
variable.
2. Create BL DS from
original variables.
Select the Raw data as
specified in the
workset of the active
model check box.
A page for selecting
variables for each phase.
See Selecting BEM
variables for BL DS
section. Finish creates
the dataset.
The variables are aligned
using the time/maturity
variable.
3. Create BL DS from
statistics of the
original variables.
Select the Raw data
statistics as specified
in the workset of the
active model check
box.
A page for selecting
which statistics to include
in the new dataset.
Finish creates the
dataset.
4. Create a BL DS
containing duration
and endpoint.
Leave the Create
duration and endpoint
in a separate dataset
check box selected.
A BL DS is created
holding the duration and
endpoint variables for all
phases in the current
BEM.
Batch
287
Options Action Result after
Next/Finish
5 For missing phases,
display missing.
Alternatively display
average batch for
missing phases.
Under Advanced, leave
the Use average batch
for missing phases
check box cleared to
create a BL DS
displaying missing value
for missing phases.
Select the check box to
display average batch
values for missing
phases.
In the created BL DS,
batches with missing
phases will display
missing or average batch
depending on this option.
6. Include secondary
variable IDs from the
BE DS to the BL DS.
Available when
there are secondary
variable IDs in the
BE DS.
Under Advanced,
select the Include
secondary variable
IDs check box.
The BL DS spreadsheet
displays the secondary
variable IDs from
the BEM.
7. Specify the BL DS
name.
In the Name field, type
the desired BL DS
name.
A BL DS is created with
the specified name.
Selecting BEM variables for BL DS dialog
When clicking Raw data as specified in the workset of the active model, to
construct the BL DS with different variable selection depending on the phase, clicking
Next opens a dialog box enabling custom settings for each phase. Projects without
phases are treated as having one phase.

User Guide to SIMCA
288
The original variables in the model and those excluded are available for selection for
each phase.
Variable Description Comment
Raw Data Original variables. Click the + to select a subset of
the original variables in the
phase.
Raw Data for
Excluded
Variables
Original variables not
included.
Click the + to select a subset of
the original variables not
included in the phase.
Select/clear the desired check boxes and click Next/Finish.
Raw Data Statistics Types dialog
When selecting Raw data statistics as specified in the workset of the active model,
clicking Next opens a dialog box for selecting which statistical variables to create in
the new BL DS.

The statistical variables, listed in the table, are calculated for each variable in each
phase.
Name in Raw
Data Statistics
page
Name in BL DS
spreadsheet
Description
Min MIN Minimum value.
Max MAX Maximum value.
Mean AVG Average value.
Median MEDIAN Median value.
Std. Dev. SD Standard deviation.
Robust Std. Dev. RSD RSD = IQR/1.075
Batch
289
Name in Raw
Data Statistics
page
Name in BL DS
spreadsheet
Description
Interquartile IQR IQR = IQR[1] - IQR[3]. That is,
the third interquartile
subtracted from the first
interquartile.
Slope SLOPE Slope of the variable.
Select the desired check boxes and click Finish.
Batch condition datasets
Batch condition variables are variables pertaining to the whole batch used in BLM.
These may be starting conditions or final (result) conditions.
A batch condition variable can be a continuous variable or discrete (qualitative).
If batch conditions were imported in the BE DS, they are available in a separate batch
level DS (batch conditions dataset, BC DS) created at import. For how to specify batch
conditions during import of the BE DS, see the Specifying data properties subsection
in the SIMCA import chapter.
For how to import batch conditions separately, see the Importing batch conditions
subsection in the SIMCA import chapter.
Batch level datasets and missing values
When creating batch level datasets, BL DS, from original variables, and there are
missing values, SIMCA interpolates all missing values. Missing values can still be
present first or last, as no extrapolation is done.
When batches are missing phases, this is usually due to the fact that these phases/steps
are not needed for those particular batches. In SIMCA there are two options:
- Replacing these phases/steps with missing when creating the BL DS. This is
the default in SIMCA. Note that this may cause bias in the results.
- Replacing the missing phases/steps by the average values of these
phases/steps in the workset (batches that yielded good results). This is the
way SIMCA treated missing phases pre 13 and by some viewed as the best
way to ensure that these phases will not bias unfavorably the predicted
quality.
User Guide to SIMCA
290
Sources of Variation plot
In batch level models it is useful to display the contribution and loading plots as line
plots over time rather than column plots at every time point. Therefore the Sources of
Variation plot is the default loading and contribution plot for all BLM with one or
more score or raw variables. The Sources of Variation plot contains the exact same
data as a loading or contribution plot would show. Instead of having the variable
number on the x-axis, the maturity of the batch is used. This gives a better view of how
the process variables relate to each other at different stages in the process.
For projects with phases, this plot is displayed showing all phases.
BLM with batch conditions
Variables without maturity, i.e. batch conditions, cannot be shown in Sources of
Variation plots. When the BLM contains batch conditions it can be useful to view the
plot as a regular column plot too.
To switch the Sources of Variation plot to a column plot, on the Tools tab, in the
Change Type group, click Column. A normal loading plot can always be created from
the Loading plot gallery, in the Diagnostics & Interpretation group on the Home tab,
by selecting one of the standard options.
Contribution plots in column form may be selected on the Marked Items tab, in the
Drill Down group, by clicking Column in the comparison plot gallery.
Batch
291
BLM loading and contribution plots example
How to change between a Sources of Variation and a column plot is described in the
table. This example uses a loading plot, but is applicable for contribution plots as well.
Step Illustration
1. On the Home tab, click
Loadings.

This plot displays how the selected variables
variation over time differs from the average of
batches or from another batch.

2. To transform to the
column plot, on the Tools
tab, in the Change Type
group, click Column.

3. To transform to the Sources of Variation plot, on the Tools tab, in the Change
Type group, click Sources of Variation.
4. To switch phases or displayed variables in the contribution plot open Properties
and:
- click the desired phase in the Select phase box.
- add to the Selected list the variables to display.
User Guide to SIMCA
292
Time/Maturity group plots
The two plots Observed vs. Smoothed Y and Unaligned vs. Aligned are found in the
Time/Maturity group on the Batch tab.

Observed vs. Smoothed Y plot
The Observed vs. Smoothed Y plot displays the original maturity variable alongside
the maturity variable as configured in the workset. This plot is displayed for each phase
and is very illustrative when the maturity variable has been smoothed.
To open the Observed vs. Smoothed Y plot, click it in the Time/Maturity group on
the Batch tab.

Use Properties to select a different batch.

Batch
293
Unaligned vs. Aligned plot
To visualize the alignment of batches, i.e., how much a batch was stretched or shrunk
to the median length of the workset batches:
1. On the Batch tab, in the Time/Maturity group click Unaligned vs. Aligned
Plot.
2. In the dialog select the model in the Data box, vector in the Item box, and
batch in the Batch box, and when applicable variable and component. Click
OK.

The resulting plot displays how well the aligned line follows the unaligned line.

User Guide to SIMCA
294
Creating hierarchical models for batch level datasets
When there are batch level datasets (BL DS) present, hierarchical batch models can
automatically be created by clicking Create Hierarchical Batch Models in the
Dataset group on the Batch tab.

The types of hierarchical base models possible are:
- One model for each phase and component available when the variables in
the BL DS are scores and the BEM has phases.
- One model for each component available when the variables in the BL DS
are scores but the BEM has no phases.
- One model for each phase in the BEM available when the BEM has phases.
- Sequential models covering a part of completion of the batches available
when the BEM does not have phases.
When there are more than one BL DS, a dialog allowing you to select which datasets to
use is opened.
Note: You can only select datasets created from the same BEM.
Create Hierarchical Batch Models dialog
The Create Hierarchical Batch Models dialog has two interfaces, one where the
BEM that the BL DS was created from has 2 or more phases and one where the BEM
has one or no phases.
The interface with one or no phases:

Batch
295
The interface with two or more phases:

When creating the hierarchical batch models, the following options are always
available:
- Calculate at least x component(s) for each base model. This option forces
extraction of x components when possible.
- Center/UV scale all score variables in base and top level models. The
default is Center. Selecting UV scale scales all score variables in the base
model, if there are score variables, to unit variance, leaving all other
variables Center scaled. Also all top level model score variables are scaled
to unit variance when UV scale is selected.
Note: The selection of Center or UV scale only affects the default scaling of
the score variables. All other variables are default scaled to unit variance.
User Guide to SIMCA
296
Create Hierarchical Batch Models results
The resulting models from a few examples of using Create Hierarchical Batch
Models are described in the table:
Characteristics Dialog settings Description of
resulting models
1.
5 phases in the
BEM, each with two
components. BL DS
from scores.

10 autofitted (but at least
one component) base
models one for each
phase and component
with all variables scaled
Ctr.
The variables included in
each base model are the
score variables related to
each phase and
component.
One hierarchical top
model holding the
hierarchical score
variables scaled Ctr.
2.
5 phases in the
BEM, each with two
components. BL DS
from scores.

5 autofitted (but at least
two components) base
models one for each
phase with all variables
scaled UV.
The variables included in
each base model are the
score variables.
One hierarchical top
model holding the
hierarchical score
variables all scaled UV.
3.
5 phases in the
BEM, each with two
components. BL DS
from original data.

5 autofitted base models
one for each phase
with all variables scaled
UV.
The variables included in
each base model are the
original variables.
One hierarchical top
model holding the
hierarchical score
variables scaled Ctr.
Batch
297
Characteristics Dialog settings Description of
resulting models
4.
5 phases in BEM,
each with two
components. BL DS
from summary
variables (raw data
statistics).

5 autofitted base models
one for each phase
with all variables scaled
UV.
The variables included in
each base model are the
summary variables.
One hierarchical top
model holding the
hierarchical score
variables scaled Ctr.
5.
No phases in the
BEM, two
components. BL DS
from scores.

4 autofitted (but at least
one component) models
each covering 25% of
the batch completion
according to the
maturity.
The first model covers 0-
25%, the second 25%-
50% etc.
The base model
variables are scaled UV.
One hierarchical top
model holding the
hierarchical scores
scaled UV.
6.
No phases in the
BEM, two
components. BL DS
from scores.

2 autofitted base models
one for each
component.
The variables included in
each base model are the
score variables.
All score variables in the
base and top models are
Ctr scaled.
User Guide to SIMCA
298
Characteristics Dialog settings Description of
resulting models
7.
No phases in the
BEM, two
components. BL DS
from original data.

4 autofitted (but at least
one component) models
each covering 25% of
the batch completion
according to the
maturity.
The first model covers 0-
25%, the second 25%-
50% etc.
All variables in the base
model are UV scaled.
One hierarchical top
model holding the
hierarchical score
variables scaled Ctr.
8.
No phases in the
BEM, two
components. BL DS
from only summary
variables.
No hierarchical batch models can be created since there are:
1. No score variables.
2. No phases.
3. No components related to the batch level variables
4. No maturity variable to create sequential models.
Hierarchical top level model
All the hierarchical base models are fitted while the hierarchical top level model is
created using the scores of the base models. This last model is unfitted for editing of
the model by for instance including other variables, such as quality variables as Y.
However, selecting to include another dataset resets to default workset resulting in that
centering, if default done by SIMCA, must be redone manually.
Batch
299
Batch Variable Importance Plot
The Variable Importance Plot (also named Batch VIP), available for batch level
models created from a scores dataset, displays the overall importance of the variable
over the whole evolution on the final quality of the batch. With phases, the plot
displays the importance of a variable by phase. With a PLS model, the Batch VIP
displays one plot for each y-variable with a column per selected phase.

To open the plot:
1. On the Batch tab, in the Variable Summary group, click Variable
Importance Plot.
2. In the Batch Variable Importance dialog, select the phases for which to
display the Batch VIP. With no phases, the Batch VIP is displayed for the
BEM used to create the batch level dataset the model was built on.
3. Click OK.

Note: The Variable Importance Plot, Batch VIP, plot is only available when
scores in the scores BL DS are selected as X in the BLM.

301
Analyze
Introduction
This chapter describes all commands on the Analyze tab.
The Analyze tab displays the analysis features not common enough to end up on the
Home tab.
Content
The following commands are available on the Analyze tab:
1. Biplot to display the scores and loadings in the same plot.
2. Inner Relation to illustrate the goodness of fit.
3. S-plots to display the S-plot, S-line or SUS Plot.
4. Contribution plots to display the influence on each variable when
comparing a point or group of points with the center, another point, or a
group of points.
5. RMSECV to indicate predictive power.
6. Y-Related Profiles to display the pure profiles of the underlying spectrum
for OPLS and O2PLS models.
7. Residuals N-Plot to display the residuals on a normal probability scale.
8. Permutations to display a measure of the overfit.
9. CV-ANOVA to display the cross validated ANOVA table.
10. CV Scores to display a cross validated complement to the regular score plot.
11. HCA for cluster analysis.
12. PLS-Tree for cluster analysis highlighting subgroup formations in latent
variables.

User Guide to SIMCA
302
Biplot
The loading vectors p, c and pc and the score vector t can be displayed Correlation
Scaled leading to the vectors p(corr), c(corr), pc(corr) and t(corr). Additionally poso
and pq for OPLS and O2PLS can be displayed correlation scaled, poso(corr) and
pq(corr). The selected correlation scaled loading vector can be displayed together with
the correlation scaled score vector in a Biplot where all points end up inside the
correlation circle of radius 1.
Open the plot by clicking Biplot in the Analysis group on the Analyze tab. The Biplot
is available for all fit methods.
Note: The weights w and w*, cannot be displayed as correlations, as the u
vectors are not orthogonal.
In the Biplot here pc(corr) and t(corr) are displayed for the first and second
components.

The default coloring for the loading vector is by Terms and is found in the Color tab
in the Properties dialog. For coloring of the score vector an additional tab Score
Color is available. For details about coloring see the Coloring from Properties
subsection in the Tools tab section in the Plot and list contextual tabs chapter.
In the Properties dialog, in addition to the Label Types tab referring to the labels on
the loading vectors, there is a Score Labels tab which refers to the labels on the scores.
The default is to not display labels on the observations.
Colors in the Biplot
In the Biplot, the different points are by default colored by Terms coloring x-
variables, y-variables, cross, square, cubic, lags and observations in different colors.
The color scheme is the same as for the regular Loadings Scatter Plot.
The colors and symbols can be changed in the Format Plot dialog, Styles node.
Analyze
303
Inner Relation Plot
For a PLS model the Y space can be explored by plotting u1 vs. u2 and interpreting it
in the same way as the t1 vs. t2 plot.
When plotting a t vector vs. a u vector for the same component, the goodness of the fit
is illustrated.
Open the plot by clicking Inner Relation in the Analysis group on the Home tab.

S-plots
S-plots menu holds the three plots: S-plot, S-line and SUS-Plot.

The S-plots provide visualization of the OPLS/OPLS-DA predictive component
loading to facilitate model interpretation. The plots are only available for OPLS models
with one y-variable and OPLS-DA models with two classes.
S-plot
The S-plot
TM
is used to visualize both the covariance and the correlation structure
between the X-variables and the predictive score t[1]. Thus, the S-plot is a scatter plot
of the p[1] vs p(corr)[1] vectors of the predictive component [Wiklund, et al.,
Analytical Chemistry, 2008]. With Pareto and Ctr scaling this plot often takes the
shape of the letter 'S'. X-variables situated far out on the wings of the S combine high
model influence with high reliability and are of relevance in the search for e.g.
biomarkers that are up- or downregulated.
User Guide to SIMCA
304
In the example below, the triangle has a high p(corr) which means a very high
reliability while the square has a high model influence partly due to its high variance in
the dataset.

S-line plot
The S-line plot is tailor-made for NMR spectroscopy data. It visualizes the p(ctr)[1]
loading colored according to the absolute value of the correlation loading, p(corr)[1].
Thus, the S-line plot is conceptually similar to the STOCSY
TM
plot pioneered by the
Nicholson group at Imperial College, London, UK [Cloarec, et al., Analytical
Chemistry, 2005]. However, it should be noted that the coloring principle is slightly
different between the two plots.
The plot below displays the predictive loading in a form resembling the original
spectra, colored according to p(corr). The top end of the color scale visualizes the
NMR shifts that influence the separation of the groups.
Figure. The S-line for an OPLS-DA model.

Analyze
305
SUS-Plot
The SUS-plot is a scatter plot of the p(corr)[1] vector from two separate OPLS models
[Wiklund, et al., 2008]. If two OPLS models have similar profiles (they capture similar
relationships between the X-variables and the single Y-variable) the X-variables will
line up along the diagonal running from the lower left corner to the upper right corner.
Such variables represent the Shared structure among the two compared OPLS models.
Conversely, X-variables that are not located along this thought diagonal, e.g. in the
upper left corner, represent structures that are Unique to either of the two models
compared.
SUS is an acronym for Shared and Unique Structures.
In the plot below, in the lower left corner, the variable with plot mark triangle is a
Shared variable. It has a negative correlation of the same size in both models meaning
that it is down regulated, i.e. lower in the genetically modified samples compared with
the control in both models. The square (at 6 o'clock) is a Unique variable for M2 as the
p(corr) for M1 is 0. The opposite is true for the pentagon variable which is Unique for
M1 and has a p(corr) close to 0 for M2.

Contribution plots
Contribution plots are used to understand why an observation differs from the
others in an X score (t), in DModX, in DModY, in Hotelling's T2Range, or in the
observed vs. predicted Y plot.
The contribution plot displays the differences, in scaled units, for all the terms in
the model, between the outlying observation and the normal (or average)
observation, multiplied by the absolute value of the normalized weight.
User Guide to SIMCA
306
Contribution plots for the model can be displayed from plots or by clicking
Contribution, in the Analysis group, on the Analyze tab.

For batch level models, it is often useful to display the variation of a variable from
a selected phase in a contribution plot, as a line plot over time, rather than a
column plot at every time point. Therefore the Sources of Variation Plot is by
default displayed when creating the contribution plot from a plot when the BLM
was created from scores or original data. For more, see the Sources of Variation
Plot subsection in the Tools tab section in the Plot and list contextual tabs
chapter.
Scores/T2 Contribution plot
To create the score contribution plot using the ribbon, on the Analyze tab, in the
Analysis group, click Contribution | Scores/T2. The Contribution Scores dialog
opens:

In the dialog:
1. Select the ID to display in the box positioned top right.
2. Select the observation ID of a normal observation, a group of
observations, or AVERAGE in the From Observation list. When
selecting AVERAGE, this corresponds to a hypothetical observation
with all its elements equal to the variable means.
3. Select the observation ID of the suspect observation or group of suspect
observations in the To Observation list.
4. Select the weight parameter in the Weights box. If the point is outside in
only one dimension, select a single weight.
Analyze
307
5. Select the dimension of the scores of the outlying observation in the
Comp 1 box and, if applicable, in the Comp 2 box.
For details about the Find feature, see the Find feature in workset dialog
subsection in the Workset section in the Home chapter.

Weights available for Scores/T2 contribution
The following weights listed in the table are available in the Weights box.
Note: The absolute values for the weight vector are used.
Weight Description
Raw No weights are used. Displays the subtraction of the values in
one observation minus the other in original units.
Normalized No weights are used, but displayed in the units of the workset.
p Using the selected dimension (of the outlying observation),
the p loadings for that dimension are used as weights.
pp Using the two selected dimensions (of the outlying
observation), the corresponding linear combinations of the p
vectors are used as weights.
PRange Using all selected dimensions, the corresponding p vectors
are used as weights. Use this option for Hotellings T2Range.
w* Using the selected dimension (of the outlying observation),
the w* vector of that dimension is used as weights.
w*w* Using the two selected dimensions (of the outlying
observation), the corresponding linear combinations of the w*
vectors are used as weights.
CoeffCS The weights are the absolute values of the PLS scaled and
centered regression coefficients, for the selected y-variable,
after the selected dimension.
VIP The weights are the VIP values after the selected component.
User Guide to SIMCA
308
Weight Description
W*Range Using all selected dimensions, the corresponding w* vectors
are used as weights. Select this option for Hotellings
T2Range.
RX The weights are the square root of the fraction of the Sum of
Squares (SS) of every term explained by the model, after the
selected dimension listed as R2VX(cum). PCA only.
Hotelling's T2Range outlier
When an observation is an outlier in a Hotelling's T2Range plot, use Contribution
Scores, select AVERAGE in the From Observation list and the suspect
observation in the To Observation list. Select the weights PRange for a PC
model and W*Range for a PLS model. Select the range of the Hotellings
T2Range plot and click OK.
Note: Group contribution is not available from the Hotellings T2Range plot.
Distance to Model X Contribution plot
To create the DModX contribution plot using the ribbon, on the Analyze tab, in the
Analysis group, click Contribution | Distance to Model X. The Contribution
DModX dialog opens:

In the dialog:
1. Select the ID to display in the box positioned top right.
2. Select the observation ID of the observation(s) with a large DModX in the
list.
3. Select the weight parameter in the Weights box.
4. With weight RX, select the number of components to use. Default is to use
all components of the active model.
Analyze
309
The Contribution plot displays the scaled residuals of every term in the model, for
that observation multiplied by the absolute value of the weight parameter.

Weights available for DModX contribution
The following weights listed in the table are available in the Weights box.
Note: The absolute values for the weight vector are used.
Weight Description
Normalized No weights are used, but displayed in the units of the workset.
RX The weights are the square root of the fraction of the Sum of
Squares (SS) of every term explained by the model, after the
selected dimension listed as R2VX(cum). This is the default.
Abs(CoeffCS)

The weights are the absolute values of the PLS scaled and
centered regression coefficients, for the selected y-variable, after
the selected dimension.
VIP The weights are the VIP values after the selected component.

User Guide to SIMCA
310
Distance to Model Y Contribution plot
Create the DModY contribution plot using the ribbon, on the Analyze tab, in the
Analysis group, click Contribution | Distance to Model Y. The Contribution
DModY dialog opens:

In the dialog:
1. Select the ID to display in the box positioned top right.
2. Select the observation ID of the observation(s) with a large DModY in the
list.
3. Select the weight parameter in the Weights box.
4. With weight RY, select the number of components to use. Default is to use
all components of the active model.
The contribution plot displays the Y scaled residuals for that observation multiplied by
the absolute value of the weight parameter.

This plot is unavailable for PCA.
Analyze
311
Weights available for DModY contribution
The following weights listed in the table are available in the Weights box.
Note: The absolute values for the weight vector are used.
Weight Description
Normalized No weights are used, but displayed in the units of the workset.
RY

The weights are the square roots of the fraction of the Sum of
Squares (SS) of every Y explained by the model, after the selected
dimension listed as R2VY(cum). This is the default.
Y Predicted Contribution plot
Create the YPred contribution plot using the ribbon, on the Analyze tab, in the
Analysis group, click Contribution | Y Predicted. The YPred Contribution dialog
opens:

In the dialog:
1. Select the ID to display in the box positioned top right.
2. Select the observation ID of a normal observation, group of observations, or
AVERAGE in the From Observation list. When selecting AVERAGE, this
corresponds to a hypothetical observation with all its elements equal to the
variable means.
3. Select the observation ID of the suspect observation or group of observations
in the To Observation list.
4. Select the weight parameter in the Weights box. There is one weight:
CoeffCSRaw.
5. Select the y-variable the observation is an outlier for in the Variable box and
the number of components to use in the Comp 1 box.
User Guide to SIMCA
312
The Contribution plot displays the differences, in scaled units, for all the terms in the
model, between the deviating observation and the normal (or average) observation,
multiplied by the normalized regression coefficient (Centered and Scaled) i.e., divided
by the length of the coefficient vector (i.e. square root of sum of squares of the
coefficients for all the terms in the model.)

This plot is unavailable for PCA models.
Weights available for YPred contribution
The following weights listed in the table are available in the Weights box.
Note: The YPred contribution plot can only be created from menu. When
creating contribution plots from observed vs. predicted plots the displayed
plot is a Score contribution.
Weight Description
CoeffCSRaw

The weights are the values of the scaled and centered regression
coefficients, for the selected Y variable, after the selected
dimension.

Analyze
313
Contribution plot for hierarchical top level models
In hierarchical top models, the contribution plot refers to the top level variables,
usually scores of base models. The background is shaded to separate variables
from the different models.
For hierarchical top level models the Contribution plot displays the top level
contributions. Double-click a variable to display the individual underlying variable
contributions.

Combined Contribution plot for batch level model
The Combined Contribution plot can be created for batch level models created
from score variables, by following the steps described:
1. Open the contribution plot of interest.
2. Mark the score variables.
3. On the Marked Items tab, in the Drill down group, click Combined
Contribution.
User Guide to SIMCA
314
The plot is resolved to a sorted batch evolution contribution plot displaying
original variables.


Contribution plot from plot
Contribution plots can be created from plots displaying observations, that is,
scores, Hotellings T2Range, DModX, DModY, and Observed vs. Predicted plots.
To display a contribution plot not using the menus, use one of the following methods:
- Select the observations and click the relevant comparison plot in Drill down
group on the Marked Items tab.
- Double-click the point or column.
- Select the observations, right-click the selection, and click Create |
Contribution Plot.
Note: Group contribution is not available from the Hotellings T2Range plot.
For more about the Drill down plots, see the Create from selection and Drill down
section in the Marked Items tab section in the Plot and list contextual tabs chapter.
Analyze
315
Example of contribution plot directly from plot
In the example here, observation 208 was double-clicked in the score plot resulting in
that the contribution plot comparing observation 208 to the average was displayed.



User Guide to SIMCA
316
Coefficient Y-Related Profiles plot
When the fitted model is OPLS or O2PLS, the Y-Related Profiles plot can be
displayed. The Y-Related Profiles are the coefficients rotated so that they display
the pure profiles of the underlying constituents in X.
This plot is only available when the number of significant predictive components
is equal to the number of y-variables.

Y-Related Profiles is available in the Analysis group on the Analyze tab.
For more, see the OPLS/O2PLS - Orthogonal PLS modeling section in
Appendix A: Statistics.
Analyze
317
RMSEcv plot
RMSECV can be regarded as an intermediary to RMSEE and RMSEP, as it applies to
the workset (as does RMSEE) but indicates predictive power (as does RMSEP). The
RMSEE, RMSEP and RMSECV values are listed at the bottom of plots wherever these
parameters are applicable.
In the plot, we can see that the 3rd component results in a higher RMSEcv-value,
which is why the autofit will retain only two components.

The RMSEcv plot is available on the Analyze tab, in the Analysis group.
For more about RMSEcv, see the RMSEE, RMSEP and RMSECV subsection in
the Appendix A: Statistics chapter.

User Guide to SIMCA
318
Residual normal probability plot
The residuals are the difference between fitted and observed values and exist both
for the X block (residual matrix named E) and the Y block (residual matrix named
F).
In the Residuals N-Plot, available in the Analysis group on the Analyze tab, the
residuals for the Y block are displayed. Consequently the plot is unavailable for
PCA models.
For transformed responses, all values are, by default, in the original (back
transformed) units. To display them in transformed units, select the Transform
predictions check box in the Predictions page in Model Options or in Project
Options to apply the change to future models only.
The residuals displayed in the residuals normal probability plot can be expressed
as:
- Raw residuals The Y-residuals in original units.
- Standardized residuals The standardized residuals are the scaled Y-
residuals divided by the residual standard deviation (RSD). This is the
SIMCA default.

The Y-residuals are plotted on a cumulative normal probability scale. This plot makes
it easy to detect the following:
- Normality of the residuals If the residuals are normally distributed the
points on the probability plot follow close to a straight line.
- Outliers One or more points are below and to the right of the normal line,
or above and to the left of the normal line, usually greater than + or 4
standard deviations.
To switch to another y-variable use the Y-variable box in the Data group on the
Tools tab, or use the Properties dialog.
Analyze
319
Permutation plot
Use the Permutation plot to check the validity and the degree of overfit for the
PLS model.
Open the Permutation Plot dialog by clicking Permutations in the Validate
group on the Analyze tab.

In the dialog:
1. Select the y-variable to make the calculations for in the Select Variable
box.
2. Enter the number of permutations in the Number of permutations to
use field. Default is 20.
3. Select the Recalculate permutations check box in the case where the
calculations have been done but you want to recalculate.

When clicking OK in the dialog, the order of the y-variable is randomly
permuted the specified number of times, and separate models are fitted to all the
permuted y-variables, extracting as many components as was done with the
original Y matrix.
The Permutation plot then displays the correlation coefficient between the
original y-variable and the permuted y-variable on the x-axis versus the
cumulative R
2
and Q
2
on the Y-axis, and draws the regression line. The intercept is
a measure of the overfit.
User Guide to SIMCA
320
CV-ANOVA
The CV-ANOVA is a diagnostic tool for assessing the reliability of PLS, OPLS,
and O2PLS models.
For more see the CV-ANOVA section in Appendix A: Statistics.

CV Scores plot
The scatter plot of the cross-validated score vectors is analogous to the scatter plot of
regular score vectors, but indicates how sensitive a model is to the exclusion of an
observation of the workset.
Open the CV Scores plot to display the cross-validated complement to the regular
scores plot. Investigate the stability of single points as well as groups by displaying the
two plots side by side.
Small changes between the regular score plot and the corresponding cross-validated
score plot, indicates a model that is stable towards the inclusion or exclusion of the
observations of the workset.
In the example below, we start with the regular score plot and mark samples in each
group close to 0, that is, close to being assigned to the wrong class. In the CV Scores
plot we see that the B5_1 sample is still on the correct side of the 0-line while the
AWT samples end up on the wrong side. If and when we have samples that cross over
to the other side, this indicates that their class assignment is uncertain.


Analyze
321
HCA
The HCA dendrogram, available in the Clustering group on the Analyze tab, is a tree
diagram used for showing the clusters generated by hierarchical cluster analysis of the
data.
For more about hierarchical cluster analysis, see the Cluster Analysis (CA),
dendrograms, Hierarchical CA (HCA), PLS section in Appendix A: Statistics.

The heights of the clusters are proportional to the distance between the clusters. That
is, when the vertical lines are tall the clusters are far apart, and when they are short the
clusters are close together.
See also the PLS-Tree section later in this chapter.
Clustering of observations (matrix rows)
For the clustering of observations, the HCA can be applied to either a score matrix T
(recommended) or a data matrix X.
HCA can be started as follows:
- Click HCA, in the Clustering group on the Analyze tab, and select the
desired data series, i.e., the basis for the clustering (specific loadings or
specific observations).
- With a score scatter plot (2D or 3D) open, on the Tools tab, in the Change
Type group, click the Other plot dialog dispatch launcher. This cluster
analysis is calculated using the vectors displayed in the plot.
- With a score scatter plot (2D or 3D) open, mark points, right-click and click
Create | Plot, Dendrogram. This cluster analysis is calculated using the
vectors displayed in the plot but only the selected items.
User Guide to SIMCA
322
When clicking HCA the Hierarchical Cluster Analysis dialog opens.
With, for example, a PCA model with 6 components, the default clustering basis
consists of the 6 scores, by default added in the Selected list. In the case where the
active model was not fitted, the original x-variables are by default available in the
Selected list.

When clicking OK the cluster analysis starts. HCA uses the default similarity criterion
(Ward) and sorting principle (Size), both changeable in the HCA page in the
Properties dialog. For more, see the HCA options subsection later in this chapter.
When the calculations are finished, the resulting dendrogram is displayed.

Clicking on the Select plot items marking tool, and clicking Group Dendrogram
Clusters displays a horizontal line with which the resolution of the dendrogram is
specified, and thereby the number of clusters.
After selecting the position of the horizontal line and clicking once, the dendrogram
and all open scores and similar plots are colored according to the groupings. For more
about the marking tool, see the Group Dendrogram Clusters marking tool
subsection later in this chapter.
Analyze
323
After marking, classes can be created according to this marking. For more, see the
Creating class models from plot marking subsection later in this chapter.

The groups found by the HCA and colored in the dendrogram are colored accordingly
in all open plots displaying observations, and are also listed in the Marked Item
dockable window, opened by selecting the Marked Items check box on the View tab.
The marked items list can be copied and pasted to e.g., Excel or Word, either group-
wise or the whole list. See also the Clustering of variables (matrix columns)
subsection next.
With the colored dendrogram or the colored score plot active, clicking Create | Class
Models in the Modify Model group on the Marked Items tab, initiates one model per
class so that data can be further analyzed. For more, see the Creating class models
from plot marking subsection in the Plot and list contextual tabs chapter.

User Guide to SIMCA
324
Clustering of variables (matrix columns)
For the clustering of variables, the HCA is applied to a loading matrix P
(recommended) or a data matrix in this domain. The clustering of variables is similar to
clustering of observations. Details about coloring, marking etc, are not described here,
see the Clustering of observations (matrix rows) subsection earlier in this chapter.
See also the PLS-Tree section later in this chapter.
HCA can be started as follows:
- Click HCA, in the Clustering group on the Analyze tab, and select the
desired data series, i.e., the basis for the clustering (specific loadings or
specific observations).
- With a loading scatter plot (2D or 3D) open, on the Tools tab, in the Change
Type group, click the Other plot dialog dispatch launcher. This cluster
analysis is calculated using the vectors displayed in the plot.
- With a loading scatter plot (2D or 3D) open, mark points, right-click and
click Create | Plot, Dendrogram. This cluster analysis is calculated using
the vectors displayed in the plot but only the selected items.
When clicking HCA the Hierarchical Cluster Analysis dialog opens. To investigate
clustering in variables, in the Select data type box select Observations and Loadings
and add the series.

A PCA-model with 7 components was the active model, results in that the clustering
basis consists of the 7 loadings. Clicking OK starts the cluster analysis and results in a
dendrogram.

Analyze
325
The groups found by the HCA and colored in the dendrogram are colored accordingly
in all open plots displaying variables, and are also listed in the Marked Item dockable
window, opened by selecting the Marked Items check box on the View tab.
HCA options
In the Calculator type section the two methods to calculate the distances between
clusters are available. Ward is by default selected. Select Single linkage by clicking it.
For more about these methods, see the Cluster Analysis (CA), dendrograms,
Hierarchical CA (HCA) section in Appendix A: Statistics.
In the Sort section, the sorting methods Size, Height, and Index are listed. For more
about the Sort section, see the Clustering algorithm parameters subsection later in
this chapter.
To flip the dendrogram, select the Mirror tree check box.

See also the PLS-Tree section later in this chapter.
Group Dendrogram Clusters marking tool
The Group Dendrogram Clusters marker tool displays a slider (line) running
horizontally across the plot enabling the marking of clusters. This slider is moved up
and down by the mouse.
Adjusting it to a vertical position corresponds to the selection of as many groups as the
number of vertical line intersections. This means that by moving the slider up and
clicking the plot, the data are divided in fewer clusters and by moving it down and
clicking the plot, the data are divided in more clusters.
The current clusters are color coded.
When the Group Dendrogram Clusters is not the selected marker tool, clicking the
dendrogram plot selects all observations or variables in the clicked branch.
User Guide to SIMCA
326
Creating a PLS-DA or OPLS-DA model from plot
marking
To create a PLS-DA or OPLS-DA model from a dendrogram plot (HCA):
1. Mark in the dendrogram plot so that the desired groups are displayed in
different colors.
2. On the Marked Items tab, in the Modify Model group, click Class | Create
PLS-DA Model or Create OPLS-DA Model.
Class | Create PLS-DA Model and Create OPLS-DA Model are available when two
(or more groups) are marked in a plot displaying observations.
Creating class models from plot marking
To create classes from a dendrogram plot (HCA):
1. Mark in the dendrogram plot so that the desired groups are displayed in
different colors.
2. On the Marked Items tab, in the Modify Model group, click Class | Create
Class Models.
Class | Create Class Models is available when two (or more) groups are marked in
any plot displaying observations.
PLS-Tree
The PLS-Tree is a hierarchical clustering tool designed to highlight sub-group
formation in latent variable space. The PLS-Tree is useful for detecting subtle
clustering occurring in higher dimensional data spaces.
To read about the background to PLS-Trees and cluster analysis, see the Cluster
Analysis (CA), dendrograms, Hierarchical CA (HCA), PLS-Trees section in
Appendix A: Statistics and the HCA section earlier in this chapter.
Initializing a PLS-Tree
Before running a PLS-Tree estimation, the user must specify the data in terms of a
workset:
- X and Y,
- observations,
- transformation,
- scaling and centering,
- etc.
Note: There must be at least one Y-variable and at least one X-variable since
the models comprising the PLS-Tree are PLS models.
Analyze
327
PLS-Tree wizard - removing outliers
With a PLS model active (mark it in the project window), start the PLS-Tree by
clicking PLS-Tree in the Clustering group on the Analyze tab. Note that the PLS
model can be unfitted or previously fitted.
A wizard is started with its first page shown here. This first page allows the user to
exclude obvious outliers based on scores, DModX, and Hotellings T2.

Clustering algorithm parameters
After clicking Next, the Algorithm Parameters page appears, where the PLS-Tree
parameters are specified. The default values are chosen to be reasonable in situations
when knowledge about the given problem is scarce.
The performance of the PLS-Tree can be influenced by means of two adjustable
algorithm parameters. These parameters are called A and B and run between 0 and 1
and regulate how the PLS models are split.
A sets the balance between the score t1 and the Y. The closer the value for A is to zero,
the more weight is attributed to the score t1.
B takes into account the group size of the resulting clusters. The closer the value for B
is to zero, the less important it becomes to have equal group sizes in the dendrogram.
This means that a division along t1 is sought that minimizes the within group variation
and hence maximizes the between group differences in t1 and Y.
User Guide to SIMCA
328

Other parameters that can be adjusted are:
- Minimum number of observations in each model - by default 5. This
number has to be > 4.
- Maximum depth of the PLS-Tree - by default 4.
- Number of components in each model - by default Autofit.
For large datasets the Autofit option takes a long time, and a 1 for single component
models is recommended. Also, for large datasets the minimum size and the maximum
depth can be increased from 5 and 4 to, say, 10 and 7, respectively.
In the Sort section the sorting of the resulting dendrogram can be changed.
- Size - positions the clusters with the most observations to the right.
- Height - positions the clusters with highest bar to the right.
- Index - positions the clusters with the lowest index number to the left.
To flip the Dendrogram, select the Mirror tree check box.
PLS-Tree resulting dendrogram
Clicking Finish in the last page of the PLS-Tree wizard starts the PLS-Tree
calculations.
After a while, a tree (dendrogram), an uncolored score plot, and the PLS-Tree model
window displaying all PLS-Tree models, are opened.
The Group Dendrogram Clusters marker tool is automatically active when a
dendrogram is the active plot. Adjusting it to a vertical position corresponds to the
selection of as many groups as the number of vertical line intersections. In this
example 5 groups were selected.
Note: The heights of the clusters are by default proportional to the number of
observations in the clusters.
Analyze
329

After selecting the position of the horizontal line and clicking once, the dendrogram
and all open scores and similar plots are colored according to the colored groups. For
more about the marking tool, see the Group Dendrogram Clusters marking tool
subsection later in this chapter.
After marking, classes can be created according to this marking. For more, see the
Creating class models from plot marking subsection later in this chapter.
The Marked Items dockable window, opened by selecting the Marked Items check
box in the View tab, displays the members of the colored groups.
For the interpretation of the individual models, mark them as active (one at a time) and
investigate in the usual way.
Group Dendrogram Clusters marking tool
The Group Dendrogram Clusters marker tool displays a slider (line) running
horizontally across the plot enabling the marking of clusters. This slider is moved up
and down by the mouse.
Adjusting it to a vertical position corresponds to the selection of as many groups as the
number of vertical line intersections. This means that by moving the slider up and
clicking the plot, the data are divided in fewer clusters and by moving it down and
clicking the plot, the data are divided in more clusters.
The current clusters are color coded.
When the Group Dendrogram Clusters is not the selected marker tool, clicking the
dendrogram plot selects all observations or variables in the clicked branch.
User Guide to SIMCA
330
PLS-Tree model window and partial models
Any model can be further modified and investigated by making the model active either
in the project window or recommended the PLS Tree model window.

As always, we recommend a score plot, loading plot, and DModX plot for each
interesting model, augmented by coefficients and VIP.
Finally, any partial PLS model in a tree can be the starting point for a new tree. Hence,
a prudent strategy for a large dataset would be to first develop a tree with depth of, say,
3, then selecting the one or two most promising/interesting models as starting points
for additional levels (sub-models) of just those branches, etc. In this way the user keeps
some control of the number of models and branches, and hence avoids the
incomprehensible mess resulting from letting any clustering loose on a large dataset
with unrestricted depth of the tree.

331
Predict
Introduction
This chapter describes all commands on the Predict tab.
Use the Predict tab to:
1. Specify a prediction dataset.
2. Classify observations with respect to a model and display the result in plots
and lists.
3. Predict results for new observations with respect to a model and display the
result in plots and lists.
Note: All vectors with the suffix PS are predicted vectors.
All predictions are made using the active model and current predictionset. When a
predictionset has not been specified, SIMCA uses the first dataset as predictionset.
Content
The commands and groups available on the Predict tab are listed here.
- Specify Predictionset group: Specify, As Dataset, As Workset,
Complement Workset (Complement WS Batches for batch projects),
Class, Delete Predictionset.
- List group: Prediction List.
- Plots group: Y PS, Score PS, Hotelling's T2PS, DModX PS, Control
Charts PS, Contribution PS, Time Series PS.
- Classification group: Coomans' Plot, Classification List, Misclassification
Table

User Guide to SIMCA
332
Specify Predictionset
In the Specify Predictionset group the following buttons are available:
Specify, As Dataset, As Workset, Complement Workset, Class, and Delete
Predictionset.


Note: With a lagged variables model, the predictionset must have the same
sampling interval as the workset used to develop the model. The first
Maximum lag number of observations is incomplete and should therefore be
discarded.
To build a predictionset, by combining observations/batches from different datasets or
removing observations/batches from the predictionset, click Specify.
The Specify Predictionset dialog displays the currently selected predictionset.

Predict
333
The table lists the functionality of the page and how to use it.
Functionality Description
Source list box Select from which source the observations in the
Available observations/batches list should originate.
The available sources are:
- Datasets the currently available datasets.
- Workset complement the observations/batches
in the selected datasets not included in the workset
of the active model, that is, the observations/batches
not used to fit the model.
- Class the observations of the selected class.
Available with class models and batch models with
phases.
- As Workset the observations in the active model.
- Predictionset the currently available
predictionsets with the current predictionset default
selected.
To combine observations from different datasets, click the
Source box, click the desired dataset, and add the
observations. Repeat selecting another dataset in the
Source box.
Predictionset Name
field
Enter the name for the new predictionset in the
Predictionset Name field.
Note: Entering/leaving a name of an already existing
predictionset overwrites that predictionset without warning.
Available
observations/batches
list
In the Available observations/batches list, all
observations/batches in the currently selected dataset in
the Source box are displayed. The number of
observations/batches, in the selected dataset, is listed.
To specify the predictionset, select the
observations/batches and click the => to find them in the
Included observations/batches list.
Find feature The Find feature works as in other dialogs. For details,
see the relevant parts of the Find feature in workset
dialog section in the Workset chapter.
Included
observations/batches
list
In the Included observations/batches list, the
observations/batches of the current predictionset are
displayed. If no predictionset was selected, the
observations/batches in the first dataset are displayed.
Show Observations
check box
In a batch project, the batches are by default displayed. To
display the underlying observations, select the Show
Observations check box.
Remove-buttons After selecting observations in the Included
observations/batches list, clicking Remove removes
them from the predictionset specification. Clicking
Remove All removes all, independent of selection.
User Guide to SIMCA
334
As Dataset
To select an entire dataset as predictionset, click As Dataset and click the dataset
name. The Prediction List spreadsheet opens displaying the selected dataset with
predictions using the active model.
Clicking As Dataset displays a menu with two headers:
- Dataset where all datasets are listed.
- Predictionset where all specified predictionsets are listed. If you have not
specified any predictionsets only the default predictionset is listed here,
which is the first imported dataset.
In the predictionset the primary observation IDs/batch IDs are saved. This means that
if you have two datasets with the same primary observation IDs/batch IDs but different
variables, predictionsets created from these datasets define the same predictionset.
Note: The predictionset uses all variables for each included observation/batch
whichever dataset the observation or batch was selected from.
As Workset
To select the observations/batches included in the active model as predictionset, click
As Workset, . The Prediction List spreadsheet opens displaying the
observations/batches of the selected model with predictions using the active model.
Complement Workset
To select the observations in the selected datasets not included in the workset of the
active model as predictionset, click Complement Workset. The Prediction List
spreadsheet opens displaying the observations with predictions using the active model.
For batch evolution models, the command is named Complement WS Batches.
For instance, if you have excluded batches from the workset, you can specify them as a
predictionset by using this menu.
Class
To select the observations included in a class as predictionset, click Class, and click
the class name. The Prediction List spreadsheet opens displaying the observations of
the selected class with predictions using the active model.
This command is available for class and phase models.
Delete Predictionset
All predictionsets, but the current one, can be deleted.
Predictionsets hold references to the selected observations or batches. Deleting a
predictionset thus only removes the references and does not affect the datasets.
Predict
335
Batch predictionset
For batch projects, only batch evolution datasets, BE DS, are available to select as
predictionsets, even if the active model is a BLM.
When specifying a BE DS as predictionset and the active model is a batch level model,
BLM, the predictionset is automatically rearranged to a batch level predictionset
behind the scenes.
For batch evolution models, when the active model is a phase model, the Prediction
List displays only the batches belonging to that phase.
The predictionset for the BLM is limited to the batches and observations included in
the BEM the BLM was created from. This means that to specify batches excluded in
the BEM as predictionset, you have to mark the BEM and on the Predict tab, in the
Specify Predictionset group, click Complement WS Batches. Then mark the BLM
and open the plots on the Predict tab.
Note: The Specify Predictionset dialog by default displays batches. To view
the observations in the batches, and exclude observations, select the Show
Observations check box.
Predictionset for filtered datasets
For models built on filtered data, only original datasets are available to select as
predictionsets.
When specifying a dataset as predictionset and the active model was built from filtered
data, the predictionset is automatically filtered in the same way.
Missing values in the predictionset
The threshold of missing values is by default 50%. When the number of missing values
in an observation or a variable exceeds the specified threshold, SIMCA displays a
warning message. The threshold default can be changed in File | Project Options.
Prediction List
To open the Prediction List, click it in the List group on the Predict tab.
In this list the model membership probability (PModXPS+) and distances to the model
(DModXPS, DModXPS+) values are colored red when the observation has a
probability (PModXPS+) value smaller than the limit (default 0.05).

The columns displaying Primary ID, Secondary IDs, Y variables, and X variables are
always displayed in the Prediction List, along with the default result columns. To
select to display other columns, right-click the spreadsheet, and then click Properties.
User Guide to SIMCA
336
Properties in Prediction List
In the Properties dialog, select the prediction results to display by selecting the
corresponding check boxes.
Results for all fit methods
1. Model membership probability PModXPS+
Displayed for the last component and lists the probability that the
observation belongs to the model. With a 95% confidence level, observations
with a probability of membership less than 5% (i.e. less than 0.05) are
considered to be outliers and not belonging to the model. The PModXPS+,
DModXPS, and DModXPS+ values are colored red when the observation is
outside the critical limit.
2. Distance to model (DModXPS)
By default SIMCA displays the normalized distance to the model, i.e.,
standard deviation of the residuals divided by the pooled RSD of the model.
To change, open Project Options or Model Options. The PModXPS+,
DModXPS, and DModXPS+ values are colored red when the observation is
outside the critical limit.
3. Distance to model + (DModXPS+)
Combination of distance to the model (DModXPS) and the distance of its
score to the normal score range of the model (when its predicted score is
outside the model score range). The PModXPS+, DModXPS, and
DModXPS+ values are colored red when the observation is outside the
critical limit.
4. Predicted Scores (tPS)
The predicted scores, tPS, for all components.
5. Workset / Testset membership (Set)
A column labeling the observation as WS, the observation was part of the
workset, or TS, the observation was not used to fit the model and belongs to
the test set (predictionset).
6. Confidence interval of tPS (+ C.I. tPS[1])
7. X predicted (XVarPredPS)
A reconstructed variable as the appropriate part of TP from the
predictionset.
Results for PLS/OPLS/O2PLS models only
The results in the columns shown for PLS/OPLS/O2PLS are always back transformed
to original units independent of whether the Transform predictions and Scale
predictions check boxes are selected in Model Options.
1. Y predicted (YPredPS)
Displays the predicted response value for all responses using the last
component.
2. Observed Predicted (YVarPS YPredPS)
Displays the differences between the y-variable in the predictionset and the
predicted y-variable for all responses. YPredPS using the last component.
3. Standard error of Y (SerrLPS, SerrUPS)
Displays the standard errors for the predicted Y values. The standard errors
are displayed for all responses using the last component. For details about the
computation, see Appendix A: Statistics.
Predict
337
4. Confidence interval of Y (+ C.I. YPredPS)
Confidence interval computed from all the cross validation rounds and jack-
knifing. These confidence intervals can only be computed for predictionset
observations, i.e., observations not used to fit the model.
5. Calculate the variables from predicted tPS
For hierarchical models the predicted Y derived from the predicted scores.

Y PS plots
The plots and lists displaying predicted responses are found on the Predict tab, in the
Plots group, by clicking Y PS. The available plots and lists are: Scatter, Line,
Column, and List.
With transformed Y variables, the observed and predicted y-variables are, by default,
back transformed to original units. To display the plot or list in transformed units,
select the Transform predictions check box in Model Options and recreate the plot.
Note: The Y PS plots and lists are only available for PLS, OPLS, and O2PLS
models and their class models.
RMSEP
At the bottom of the plots, RMSEP (Root Mean Square Error of Prediction) is
displayed. This is the standard deviation of the predicted residuals (errors). It is
computed as the square root of (E(obs-pred)
2
/N).
User Guide to SIMCA
338
Y PS Scatter Plot
To display observed vs. predicted y as a scatter plot, click Y PS | Scatter.
An example of the plot follows:

Y PS Line Plot
To display the predictionset variable YVarPS and the predicted variable YPredPS
versus Num, click Y PS | Line.
An example of the plot follows:

Predict
339
Y PS Column Plot
To display Y Predicted as a column plot with confidence intervals, click Y PS |
Column.
The plot is displayed with jack-knife uncertainty bars. These limits can be modified in
the Properties dialog. For more, see the Limits subsection in the Plot and list
contextual tabs chapter.

Y Predicted List
The Y Predicted List displays for each y-variable:
Vector Description
YVarPS Measured or observed Y if present.
YPredPS Predicted Y.

Note: For discriminant models the Y Predicted List is identical to the
Classification List.
User Guide to SIMCA
340
Score PS plots
There are four types of predicted score plots available by clicking Score PS: Scatter,
Line, Column, and 3D.
These plots and dialogs are very similar to those of the score plots found on the Home
tab. For details about the dialog, see the Scores section in the Home chapter. An
example of each plot type follows here.
Predicted Score Scatter Plot
To open the scatter plot displaying predicted scores, tPS, click Score PS | Scatter.

Predicted Score Line Plot
To open the line plot displaying predicted scores, tPS, click Score PS | Line.

Predict
341
Predicted Score Column Plot
To open the column plot displaying predicted scores, tPS, click Score PS | Column.
The plot is displayed with jack-knife uncertainty bars. These limits can be modified in
the Properties dialog. For more, see the Limits subsection in Tools tab section the
Plot and list contextual tabs chapter.

Predicted Score 3D Plot
To open the 3D scatter plot displaying predicted scores, tPS, click Score PS | 3D.

For how to mark, zoom, rotate, move the plot in its window, color and size the 3D plot,
see the 3D Score Scatter Plot section in the Home chapter.
User Guide to SIMCA
342
Hotellings T2Range plot
Predicted Hotelling's T2Range can be displayed in two plot types available by clicking
Hotelling's T2PS | Line and Column.

The Hotelling's T
2
Range plot displays the distance from the origin in the model plane
(score space) for each selected observation. The plot shows the T
2
calculated for the
range of selected components, e.g., 1 to 7, or 3 to 6.
Default is to display from the first to the last component. For OPLS and O2PLS the
range is locked to from first predictive to last orthogonal in X component.
To change the range of components, right-click the plot and select Properties, tab
Component.
Values larger than the green limit are suspect (95%), and values larger than the red
limit (99%) can be considered serious outliers.
See also the Hotellings T2Range section in the Home chapter.
DModX PS
Distance to Model is an estimate of how far from the model plane, in the X or Y
space, the observation is positioned.
The distance to the model can be displayed in absolute and normalized units. By
default the distance to model plots are displayed in normalized units after the last
component.
Use Model Options or Project Options, to change units of the DMod plot.
For details about the calculations of DModXPS, DModXPS+, DModYPS and D-Crit,
see the Distance to the model section in Appendix A: Statistics.
Predict
343
DModX PS - Distance to Model X-Block
To display the DModX for the predictionset, click DModX PS and under the DModX
header click Line or Column.
The plot by default displays DModXPS+ after the selected component with limit at
significance level 0.05.
To switch to the regular DModXPS, open the Properties dialog, click the DMod tab,
and click Regular.

DModXPS+, displayed in this plot, is a combination of the regular DModXPS and the
distance to normal score range. For details, see the Distance to the model of new
observations in the predictionset subsection in Appendix A: Statistics.
DModY PS - Distance to Model Y-Block
To display the DModY for the predictionset, click DModX PS, and under the DModY
header click Line or Column.
The plot displays DModYPS after the selected component.

User Guide to SIMCA
344
Control Charts PS
With a fitted model, the predicted scores, x and y-variables and their residuals can be
displayed in control charts available by clicking Control Charts PS.

To display a control chart:
1. Select:
a. The data source in the Data box.
b. The vector to display in the Item box. tPS, XVarPS, XVarResPS,
YVarPS, and YVarResPS are available.
c. Variable and/or component in the Variable or Comp boxes when
appropriate.
2. In the Type of Control Chart box, select Shewhart, EWMA, CUSUM, or
EWMA/Shewhart.
3. Depending on the selected control chart, select the properties as desired.
4. Clicking OK opens the control chart with limits computed from the workset.
For details, see the Control charts section in the Plot/List chapter.
Predict
345
Contribution PS
Contribution plots are used to understand why an observation differs from the others in
an X score (t), in DModX, in DModY, in Hotelling's T2Range, or in the observed vs.
predicted Y plot.
The contribution plot displays the differences, in scaled units, for all the terms in
the model, between the outlying observation and the normal (or average)
observation, multiplied by the (absolute) value of the normalized weight.
Contribution plots for the predictionset can be displayed from plots or by clicking
Contributions PS.

For BLM with one or more variables created from scores or original variables in
the BEM, it is useful to display the variation of a variable from a selected phase in
a contribution plot, as a line plot over time, rather than a column plot at every time
point. For this purpose the Sources of Variation Plot is by default displayed
when creating the contribution plot from a plot. For more, see the Sources of
Variation Plot subsection in the Batch chapter.
Scores/T2
To create the score contribution plot for the predictionset using the buttons, click
Contribution PS | Scores/T2.
The Contribution Scores dialog that opens is identical to the one opened by
clicking Contribution | Scores/T2 in the Analysis group on the Analyze tab. For
more about this dialog and the available weights, see the Scores/T2 contribution
plot section in the Analyze chapter.

User Guide to SIMCA
346
Distance to Model X-Block Contribution plot
To create the DModX contribution plot for the predictionset using the menus,
click Contribution PS | Distance to Model X-Block.
The Contribution DModX dialog that opens is identical to the one opened by
clicking Contribution | Distance to Model X Block in the Analysis group on the
Analyze tab. For more about this dialog and the available weights, see the
Distance to Model X Block contribution plot section in the Analyze chapter.

Distance to Model Y-Block Contribution plot
With a PLS model, create the DModY contribution plot for the predictionset using
the menus by clicking Contribution PS | Distance to Model Y-Block.
The Contribution DModY dialog that opens is identical to the one opened by
clicking Contribution | Distance to Model Y Block in the Analysis group on the
Analyze tab. For more about this dialog and the available weights, see the
Distance to Model Y Block contribution plot section in the Analyze chapter.

Predict
347
Y Predicted Contribution plot
With a PLS model, create the YPred contribution plot using the menus by clicking
Contribution PS | Y Predicted.
The YPred Contribution dialog that opens is identical to the one opened by
clicking Contribution | Y Predicted in the Analysis group on the Analyze tab.
For more about this dialog and the available weights, see the Y Predicted
contribution plot section in the Analyze chapter.

User Guide to SIMCA
348
Contribution plot from plot
Contribution plots can be created from plots displaying observations, that is,
scores, Hotellings T2Range, DModX, DModY, and Observed vs. Predicted plots.
To display a contribution plot not using the menus, use one of the following methods:
- Select the observations and click the relevant comparison plot in Drill down
group on the Marked Items tab.
- Double-click the point or column.
- Select the observations, right-click the selection, and click Create |
Contribution Plot.
Note: Group contribution is not available from the Hotellings T2Range plot.
For more about the Drill down plots, see the Create from selection and Drill down
section in the Marked Items tab section in the Plot and list contextual tabs chapter.
Example of contribution plot directly from plot
In the example here, observation 208 was double-clicked in the score plot resulting in
that the contribution plot comparing observation 208 to the average was displayed.



Predict
349
Time Series PS
The Time Series PS plots, display selected prediction vectors as series on the Y-axis
and Num on the X-axis. The differences versus using the Plot/List | Line Plot is that
Num, or Date/Time when available, is always plotted on the X-axis and the vectors
available in the Item box are the prediction vectors only.
To open a time series plot, click Time Series PS in the Plots group on the Predict tab
(or Time Series in the Control Charts group on the Plot/List tab).
In the Time Series Plot dialog, select the vectors to plot in the Item box and click the
Add Series-button.
To display all vectors scaled to range 0 1, select the Scale 0-1 check box.

Time Series plot example
The plot here displays the observed y-variable in the predictionset, YVarPS, and the
predicted y-variable, YPredPS.

User Guide to SIMCA
350
Coomans' Plot
Use Coomans' plot to classify new observations with respect to two selected models.
For each observation in the predictionset, the distance to each of the selected models is
computed and plotted along with the selected models critical distances.
To display Coomans' Plot:
1. Specify the predictionset.
2. On the Predict tab, in the Classification group, click Coomans' plot.
3. In the Coomans' Plot dialog, select the two models to plot the DModXPS+
for.
4. Optionally click the Color tab and select to color by classes.
5. Click OK.

In the Limits page you can select to hide the critical lines. In the DMod page you can
select to display regular DMod. The other tabs are general and described in detail in
the Properties dialog section in the Plot and list contextual tabs chapter.

Predict
351
Classification List for class models
With class models you can automatically classify a predictionset with respect to all the
class models by:
1. clicking Classification List in the Classification group on the Predict tab,
2. selecting the desired models, and clicking OK.

The Classification List displays the observations of the predictionset as rows, and the
class models probability of membership, PModXPS+, as columns. The model title in
the second row identifies the class.
Note: By default the primary observation ID is displayed. To display all
observation IDs, on the Tools tab, in the Properties group, click Labels. For
more see the Label Types for lists subsection in the Tools tab section in the
Plot and list contextual tabs chapter.
Coloring in the Classification List
The values displayed in the Classification List are probabilities.
The observations are colored such that probabilities:
- > 0.10 are green (inside 90% confidence of the normal probability curve).
- between 0.10 and 0.05 are orange (inside 95% confidence).
- < 0.05 are white and the observation is deemed to be outside the class
(outside 95% confidence).

User Guide to SIMCA
352
Classification List for Discriminant Analysis models
The Classification List for discriminant analysis models provides the predicted Y
value for the dummy variables (0 or 1) used to direct the projection.
The Classification List displays the observations of the predictionset as rows, and the
original dummy variable in YVarPS and predicted dummy variable in YPredPS, as
columns.
Note: By default all observation IDs are displayed. To different IDs, on the
Tools tab, in the Properties group, click Labels. For more see the Label
Types for lists subsection in the Tools tab section in the Plot and list
contextual tabs chapter.
Coloring in the Classification List
The values displayed in the Classification List for discriminant analysis models are
the predictionset original values and predicted values.
The observations are colored such that predicted values:
- < 0.35 are white (do not belong to the class).
- between 0.35 and 0.65 are orange (borderline).
- above 0.65 are green (belong to the class).

Membership of a class depends upon matching the value of the dummy variable, so a
value close to one indicates membership of the workset class. In practice 0.5 is often
used as a practical threshold in order to classify an observation as belonging to one
class or another.
Misclassification Table
The Misclassification Table is available for all class models and discriminant analysis
models.
The Misclassification Table shows the proportion of correctly classified observations
in the predictionset.
If the class information is not available in the dataset that the predictionset was created
from, the class information has to be defined. This is achieved by:
- Using the Dataset Properties dialog, Observations page. See the
Observations page subsection in Dataset Summary section in the Data
chapter.
- Specifying a Class ID while importing the file. See Class ID specification
subsection in the Home tab section in the SIMCA import chapter.
Predict
353
Misclassification Table for class models
The Misclassification Table is available for all class models.
With specified classes, the Misclassification Table summarizes how well the selected
models classify the observations into the known classes. By default the Assign each
observation only to the nearest class option is used (see the Properties dialog
described in the Misclassification Table Properties subsection later in this section).

The Misclassification Table is organized as follows:
Column Content
First column The class names, No class, Total, and Fishers probability.
The Total row displays: the total number of observations in the
predictionset, the average percentage correctly classified, and
the number of observations classified to each class.
Fishers probability is the probability of the table occurring by
chance and is satisfied when p < 0.05 for 95% confidence. For
more, see the Fisher's Exact test subsection in Appendix A:
Statistics.
Fishers probability is displayed when Assign each observation
only to the nearest class is the selected option in Properties.
Members The number of observations assigned to each class in the
selected predictionset. At the bottom of the list Fisher's
probability value is listed.
Correct The percentage correctly classified for each class and at the
bottom the average percentage.
Individual
classes
The number of observations classified as belonging to the class
on that row.
The correct classifications are colored green while the incorrect
are colored yellow.
No class
(PModX+<'limit')
The observations that are not classified to any one of the
available classes.
Combined
classes
The observations that overlap between classes.
When Assign all observations to classes above the limit is
selected, all combinations of classes are displayed with number
of overlapping observations.
Note: All entries on the diagonal are observations that have been correctly
classified.
User Guide to SIMCA
354
In the above table, we can see that all the observations but 6 observations were
correctly classified, one belonging to Ver was classified as Vir, and 5 belonging to Vir
were classified as Ver. The reason for this is the closeness of the two classes,
illustrated in the score plot here.

Misclassification Table - Discriminant Analysis models
The Misclassification Table is available for all discriminant analysis models.
By default the observations are assigned to the nearest class only, using no threshold.
To see the options available, see the Misclassification Table Properties subsection
later in this section.

For a description of the content of the table, see the Misclassification Table for class
models subsection earlier in this section.
Misclassification Table Properties
The Misclassification Table Properties dialog contains options to change the way of
classifying the observations.
Use the Properties dialog to:
- Select models in the Model(s) box.
- Select how to classify the observations and probability level according to the
method below.
Predict
355
Method for assigning observations
Select whether to assign each observation:
1. Only to the nearest class, with an optional PModX cut-off by clicking Assign
each observation only to the nearest class.
2. To all classes if they exceed the probability limit listed by clicking Assign
observations to all classes above the limit. When there are overlapping
classes, or classes close to each other, this may result in that observations are
assigned to several classes.
In the Do not assign observations with PModX below 'limit' to any class field,
optionally enter the probability limit to use.
A possible reason for entering a probability limit is in cases of unequally sized groups
where the discriminating threshold may not be in the center of the projection.
The default values in this field are:
- '0' for Assign each observation only to the nearest class always.
- '0.1' for Assign observations to all classes above the limit, when the
models are class models
- '0.65' for Assign observations to all classes above the limit, when the
model is a DA-model. Another common value is '0.5'.
Note: The fields remember the last entered value.


357
Plot/List
Introduction
This chapter describes the Plot/List tab features.
The Plot/List tab allows plotting and listing input data such as observations and
variable values, computed elements such as scaling weights, variable variances, etc, as
well as results such as loadings, scores, predictions, etc, of all the fitted models.
The plot types available on the Plot/List tab are:
- The Standard Plots and list: Scatter, Scatter 3D, Line, Column, and List.
- The Control Charts: Time Series and Control Charts.
- The Custom plots: Histogram, Normal Probability, Response Contour,
Response Surface, Wavelet Structure, Wavelet Power Spectrum and
Step Response Plot.
The plot types in the Standard Plots group are not described in this user guide, but the
dialog interface is described in this chapter.
Content
This chapter describes the following plots Control Charts, Response Contour,
Response Surface, Wavelet Structure, Wavelet Power Spectrum, and the Step
Response Plot.

Note: Using the Change Type group on the Tools tab, plots can be created
from plots and lists, and lists can be created from plots and lists.
Name conventions
Some vectors have letters appended to them at the end of the item (vector) name.
These indicate the source of the data, e.g. intermediate model results from cross
validation rounds, statistics computed from all the cross validation rounds, or other
information about the presented vector.
Note: Vectors containing Y are available for PLS, OPLS, and O2PLS models
but unavailable for PCA models.
User Guide to SIMCA
358
See the table for the added endings and their description.
Ending Description
DS Any DataSet. Variables or observations in datasets are always the
original raw variables with no role definition, or treatment. To access
them, select the appropriate dataset in the Data box in the Plot/List
dialog.
PS The current PredictionSet. There is only one current predictionset at
any time. Variables or observations in predictionsets are always the
original raw variables with no role definition, or treatment.
All items with the two letters PS are available when selecting a model in
the Data box. All items without the two source letters are part of the
selected model i.e., they pertain to variables or observations in the
workset associated with the model.
To plot or list variables from the predictionset directly, select the
predictionset as data source in the Data box.
cv Intermediate results such as scores, loadings, PLS weights, predicted
values, etc., computed from a selected cross validation round.
cvSE Jack-knife standard error computed for model results from all the cross
validation rounds.
St The vector is in standardized units, i.e. divided by its standard
deviation.
(trans) The vector is in transformed units according to the workset
specification.
(as WS) The vector is displayed in the same units as the workset, WS. This
means that a variable scaled to unit variance in the workset is displayed
in scaled and centered units, while a transformed variable is displayed
in the transformed metric, scaled and centered.
Cum The vector is cumulative over all components up to the listed one.
Note: There are a number of vectors that are cumulative without
displaying this ending.
OOCSum The vector is an Out Of Control summary vector.
Standard plot types and list dialog
The standard plot types available on the Plot/List tab are Scatter, Scatter 3D, Line,
Column, and List.
A selection of the following pages is available in different Plot/List dialogs: Data
Series, Label Types, Item Selection, Color, Transformation, Size, Limits, and
Number Format. The Data Series and Transformation pages are described in the
subsections that follow as they are unique for the Plot/List dialogs.
For more about the other pages, see the Properties dialog subsection in Tools tab
section in the Plot and list contextual tabs chapter.
Plot/List
359
Data Series page
In the Data Series page the vectors (series) to be displayed in the selected plot or list
can be selected. To make the vector available in the Item box, the appropriate data
type and data source have to be selected. Some vectors require selecting a variable,
component(s), or CV group and some can be displayed in transformed and scaled units.
The list of currently added series is displayed in the Selected list. To remove series
from the list, select them and click the Remove button, or click the Remove All
button.
By selecting the Scale 0-1 check box all series are scaled so that the maximum value is
1 and minimum is 0.

Note: When the dataset contains numerical IDs, they can be selected and
plotted on any axis.
Select data type
Before selecting the data series for plotting, the appropriate data type has to be selected
from the Select data type box.
The available data types are:
- Variables and Scores vectors with one element per observation.
- Observations and Loadings vectors with one element per variable.
- Function of component, F(Component) vectors that apply per component.
- Function of Lags, F(Lags) vectors specific to a lagged model, such as
pLag.
- Aligned Vectors aligned vectors available in batch control charts for batch
evolution models.
- Batch Vectors Out Of Control (OOC) variables for the aligned vectors for
batch evolution models.
User Guide to SIMCA
360
Data source
By default the active model is the selected data source. For computed elements such as
scaled variable values, variable variances, model results, or intermediate results from
cross validation, the data source can only be a fitted model, not a dataset.
The data sources available in the Data box are:
1. DS1 dataset first imported.
2. DS2, DS3 etc. other datasets.
3. PS current predictionset.
4. M1, M2 etc. fitted models.
To access items specific to another data source than the default, select the data source
in the Data box.
For instance, for datasets preprocessed using Orthogonal Signal Correction (OSC), the
OSC scores and loadings are available after selecting the OSCed dataset in the Data
box.
Item
In the Plot/List tab dialogs, all items from the selected data type and source are
available in the Item box.
Some vectors require selecting:
- Variable from the Variable box.
- Dimension from the Component box for some model results.
- Cross validation round from the CV box for intermediate results from cross
validation.
Some items, such as DModX, are available for component 0, which correspond to
scaled and centered data as specified in the workset (WS).
For Hotelling's T2Range specifying the starting and ending components is required.
Note: Vectors containing Y are available for PLS, OPLS, and O2PLS models
but unavailable for PCA models.
For a description of all vectors available for the Plot/List tab plot types, see the
Vectors available in SIMCA section in Appendix A: Statistics.
Scale and Transform
When there are transformed variables in the model, many vectors can be displayed in
the transformed metric. These same vectors can always be displayed scaled as the
workset. Examples of such vectors are XVar and XVarRes.
- Select the Transform check box, before clicking Add Series, to display the
vector in the transformed metric. Or after adding the vector, in the Selected
list, select Yes in the Transform column on the row of the vector.
- Select the Scale check box before clicking Add Series, to display the vector
in the metric of the workset. Or after adding the vector, in the Selected list,
select Yes in the Scale column on the row of the vector.
Plot/List
361
For example, with Scale selected, a variable scaled to unit variance in the workset is
displayed in scaled and centered units. If the variable was transformed, it is displayed
in the transformed metric, scaled and centered. Hence, when selecting the Scale check
box, the Transform check box is automatically selected.

Note: When both the Scale and Transform boxes are cleared, the added
vector is displayed in original (back-transformed) units.
Scaling and Offset
Specifying a Scaling value or Offset value is useful with Line and Time Series Plots
when the selected vectors need to be visually separated.

After adding the series, enter values in the Scaling and/or Offset columns in the
Selected list.
Transformation
On the Transformation page the following data transformations are available:
- Auto Correlation
- Cross Correlation
- Power Spectrum
- Wavelet Coefficients
- EWMA
- Histogram
- Normalize
- R2X
User Guide to SIMCA
362

Auto or cross correlation of variables or observations
The auto and cross correlation are measures of dependence of adjacent observations
and characterizes a time series or observation profile, in the time domain.
For technical details, see the Auto and cross correlation of variables or observations
section in Appendix B: Preprocessing.
Spectral filtered wavelet data
With a dataset that has been wavelet transformed and compressed variable wise (using
Spectral Filters on the Data tab), the auto/cross correlation observation wise refers to
the reconstructed observations, when the Reconstruct wavelets check box is selected.
To display the auto/cross correlation in the wavelet domain clear the Reconstruct
wavelets check box found in Project Options on the File tab.
Applying Auto Correlation
To apply auto correlation to a variable or observation:
1. Select the desired vector on the Data Series page.
2. Click the Transformation tab.
3. Select Auto Correlation in the Select transformation box.
4. In the Detrend box the default is Mean. Select
a. Linear to remove the best linear fit, or
b. None to do nothing, or
c. leave Mean to remove the mean.
5. To change maximum lagging, type a new value in the Max Lag field. The
auto correlation is default max lag L=30 or N/4, whichever is smaller, and
can be displayed up to L=N.
Plot/List
363

Applying Cross Correlation
To apply cross correlation to a variable or observation:
1. Select the desired vector on the Data Series page.
2. Click the Transformation tab.
3. Select Cross Correlation in the Select transformation box.
4. Select the common vector to use cross correlation with by selecting data
source in the Data box, vector in the Item box, and if applicable, variable in
the Variable box, component in the Comp box, and cross validation group
in the CV box.
5. In the Detrend box the default is Mean. Select
a. Linear to remove the best linear fit, or
b. None to do nothing, or
c. leave Mean to remove the mean.
6. To change maximum lagging, type a new value in the Max Lag field. The
auto correlation is default max lag L=30 or N/4, whichever is smaller, and
can be displayed up to L=N. The plot displays the cross correlation for L to
+L.

User Guide to SIMCA
364
Power Spectrum Density of variables or observations
The power spectrum density (PSD) is the representation of the sequence x (t) in the
frequency domain. The power spectrum is the square of the amplitude of the Fourier
component at each frequency.
The frequency domain representation is particularly helpful to detect irregular cycles
and pseudo periodic behavior, i.e. tendency towards cyclic movements centered on a
particular frequency.
SIMCA uses the Welschs non-parametric method to estimate the PSD.
For more information see the Power Spectrum Density section in Appendix B:
Preprocessing.
Spectral filtered wavelet data
With a dataset that has been wavelet transformed and compressed variable wise (using
Spectral Filters on the Data tab), the PSD observation wise refers to the reconstructed
observations, when the Reconstruct wavelets check box is selected.
To display the PSD in the wavelet domain, clear the Reconstruct wavelets check box
found in Project Options on the File tab.
Applying Power Spectrum
To apply the Power Spectrum to an observation or a variable, either use the Plot/List
dialog described here, or use the Quick Info:
1. Select the desired vector on the Data Series page.
2. Click the Transformation tab.
3. Select Power Spectrum in the Select transformation box.
Change the options as warranted and click OK to create the plot. For more about these
options see the Power Spectrum Options tab subsection in the Quick Info section in
the View chapter.

Plot/List
365
Wavelet Coefficients
The wavelet transform of a signal X(t) is the representation of the signal in both the
time and frequency domain. It is the scalar product between the signal x(t) and a
mother wavelet function, stretched or compressed to create different scales (inverse of
frequency), changing the window width.
Because of the repeated re-scaling, the variable or observation is decomposed into its
details at every scale; these are the wavelet transform coefficients.
SIMCA uses the Discrete Wavelet transform, with orthogonal or biorthogonal wavelets
and the fast Multi Resolution Analysis algorithm of S. Mallat.
Select the variables or observations you want to wavelet transform and the range of
observations or variables to include.
Applying Wavelet Coefficients
To apply the wavelet coefficients to an observation or a variable, either use the
Plot/List dialog described here, or Spectral Filters in the Filter group on the Data
tab:
1. Select the desired vector on the Data Series page.
2. Click the Transformation tab.
3. Select Wavelet Coefficients in the Select transformation box.
4. Change from the default detrending, wavelet function, and wavelet order if
warranted. For more details about these options, see Wavelet
transformations section in Appendix B: Preprocessing.

User Guide to SIMCA
366
EWMA
The EWMA transformation can be used to smooth scores or variables in plots.
As an example the following line score plot of t1 vs. t2 was transformed by EWMA
and selecting subgroup 10.


The EWMA transformed plot is easier to interpret.

Plot/List
367
Applying the EWMA transformation
To apply the EWMA to an observation or a variable:
1. Select the desired vector on the Data Series page.
2. Click the Transformation tab.
3. Select EWMA in the Select transformation box.
4. Enter the sample size in the Sample Size field.
5. Select User Entered in the Lambda box and enter a value in the field, or
leave Estimated.
6. Select User Entered in the Target box and enter a value in the field, or
leave Estimated.
7. Select User Entered in the Standard Dev. box and enter a value in the field,
or leave Estimated.
8. Click OK.
Histogram
To transform an observation or a variable to a histogram:
1. Select the desired vector on the Data Series page.
2. Click the Transformation tab.
3. Select Histogram in the Select transformation box. Click OK.
Note: Histogram is available in the Transformation page when only one
series was added.
Histograms can also be created by clicking Histogram in the Custom Plots group on
the Plot/List tab.
Normalize in Transformation page
OPLS and O2PLS loading plots are by default normalized to unit length (Normalize to
unit length check box is selected in Properties). When creating loading plots on
Plot/List tab, the Transformation page must be used.
The Normalize transformation is only available for the vectors p, po, q, qo, so, and r
when the model type is OPLS and O2PLS.
The normalization will make the sum of squared vector elements equal to 1. One
advantage with this is that the loading line plots of spectral data are much easier to
interpret. Normalize transformed loading plots should be used in conjunction with R2X
transformed score plots.
R2X in Transformation page
OPLS and O2PLS score plots are by default scaled proportionally to R2X (Scale
proportionally to R2X check box is selected in Properties). When creating score
plots on Plot/List tab, the Transformation page must be used.
The R2X transformation is only available for the vectors t, to, u, uo, tPS and toPS
when the model type is OPLS and O2PLS.
The R2X transformation will make distances in the score plot correspond with the
explained variation (R2X). R2X transformed score plots should be used in conjunction
with normalized loading plots.
User Guide to SIMCA
368

Control Charts
Control Charts are available on the Predict and Plot/List tabs.
The control charts from the Predict tab, will only display items from the predictionset.
Opening the control chart from the Plot/List tab includes both the predictionset and
workset vectors (items).
The Control Chart group on the Plot/List tab holds both the Control Chart and the
Time Series buttons whilst this section only documents the Control Chart. For more
about the Time Series plot, see the Time Series PS section in the Predict chapter.
What is a control chart?
Control charts, also called statistical process control charts, constitute one of the major
tools of Total Quality Management. They enable displaying the data measured on the
process (or product) over time, and thereby detect process upsets, shifts, trends, etc.,
thus helping to improve the quality of the products.
However, processes generate collinear multivariate data. Therefore, monitoring and
charting each variable as if it is independent from the others (univariate control charts)
may be highly misleading, and may actually lead to the degradation of the product
quality. Multivariate control charts use all the process variables simultaneously, and
also extract the information contained in their correlation structure. This allows
tracking of the process over time, identify process points and regions where the process
is operating normally, as well as when the process starts to go out of control.
Because the scores (T and U) are linear combinations of the original variables, they are
close to normally distributed as long as the process is stable. The scores are also much
less noisy than the original variables since they have the characteristics of weighted
averages. Hence, multivariate control charts are more reliable than univariate charts,
and in addition much more informative.
Multivariate control charts
Multivariate control charts are model dependent, because the combination of all the
variables to summaries are made via a multivariate model, here PCA, PLS, OPLS or
O2PLS.
PC models
With a principal components (PC) model, the scores (T) are the best summary of all
the process variables. Using the score values of one model dimension (t1, t2, etc.) as
the control chart variable, yields a multivariate control chart based on all the process
variables. The chart shows the position of the process along the direction of the
specified model dimension. This chart can be used to track and follow the process as
summarized by that dimension.
PLS models
With a PLS model, the X-scores (T) provide the best summary of the process variables
that dominate the product qualities, and the Y-scores (U) provide the best summary of
the product quality. Control charting the t-vectors or the u-vectors allows tracking the
process using all the data.
Plot/List
369
OPLS and O2PLS models
For OPLS and O2PLS models, the X-scores (T) provide the best summary of the
process variables that dominate the product qualities, and the Y-scores (U) provide the
best summary of the product quality. Control charting the t-vectors or the u-vectors
allows tracking the process using all the data.
In addition, the charting of To and Uo open up for analysis of the orthogonal
information.
Univariate control charts
Univariate control charts are model independent, and can be plotted for a specified
dataset.
To display a univariate control chart with some observations excluded: create an
appropriate model. Control charting a variable from the model displays only the
observations of that model's workset.
Note: To look at special details of a variable, create control charts of an
individual variable.
Control Chart dialog
To open the Control Chart dialog, click Control Chart PS in the Plots group on the
Predict tab or Control Chart in the Control Charts group on the Plot/List tab.

Types of control charts
There are four types of control charts available in the Type of Control Chart box:
Shewhart, EWMA, CUSUM, and EWMA/Shewhart.

User Guide to SIMCA
370
Selecting the vector to display in control chart
To select the vector (item) to display in the control chart:
Step Description Action Screen shot
1

In the Data box, the available
data sources are listed. The
data sources are the models
(Mxx), datasets (DS1, DS2,
etc) and the current
predictionset (PS).
Select the
data source
in the Data
list box.

2 In the Item box, the available
items are listed: t, tPS, u,
XVar, XVarPS, XVarRes,
XVarResPS, YVar, YVarPS,
YVarRes, and YVarResPS.
Select the
vector in the
Item box.

3 For univariate control charts
the variables are listed in the
Variable/X Terms/Y Terms
box. For the X terms the
expanded terms are also
listed.
Select the
variable in
the
Variable/X
Terms/Y
Terms box.

4 In the Comp box, the
components for the selected
vector are listed when the
vector is a score or residual
vector.
Select the
component
to display the
vector for in
the Comp
box.
Note: To display multivariate control charts of t or u, and for items from the
predictionset, you must select a model in the Data box.
Plot/List
371
Shewhart
Three types of Shewhart charts are available in SIMCA:
- Individual (default) where the sample size is 1.
- Mean/Range where the default sample size is 5 and can be changed as
desired.
- Mean/StdDev where the default sample size is 5 and can be changed as
desired.

Individual
To create the Shewhart control chart without subgroups:
1. Click Control Chart.
2. Select Shewhart in the Type of Control Chart box.
3. Leave Individual (the default) selected in the Shewhart Type box, and then
click OK.
The resulting plot displays the Shewhart control chart with values for each
observation of the selected vector, with the 2 and 3 sigma limits displayed.
When plotting t or tPS, a plot of DModX or DModXPS is displayed beneath, with its
critical distance. Observations outside the critical distance indicate process behavior
different from the model.

The plot footer displays: S(Mxx), S(UE) when defined, Target(Mxx) or Target(UE)
when defined, R2X[a], 1 R2X(cum)[last component] and Mxx D-Crit[last
component].
User Guide to SIMCA
372
To view the nomenclature and calculations of target and standard deviation, see the
Nomenclature and notation and Target and Standard deviation subsections in
Appendix A: Statistics.
Mean/Range
To create the Shewhart control chart of type Mean/Range:
1. Click Control Chart.
2. Select Shewhart in the Type of Control Chart box
3. Select Mean/Range in the Shewhart Type box.
4. Enter the desired sample size in the Sample size field (default 5), and then
click OK.
The resulting plot displays the control charts for the selected vector, with the target and
upper and lower limits displayed.

The plot footer displays: Target(Mxx) or Target(UE) when defined, LCL(x),
UCL(x), S(Mxx), S(RAvg) or S(UE) when defined, RAvg(within, Mxx or S(UE)
when defined), LCL(r) and UCL(r).
To view the nomenclature and calculations of target and standard deviation, see the
Nomenclature and notation and Target and Standard deviation subsections in
Appendix A: Statistics.
Plot/List
373
Control limits and process range
Control limits and estimates of the process range in the Shewhart Mean/Range
control chart are computed as follows:
- S(RAvg) = RAvg /d2, estimate of process standard deviation.
- LCLx = Target - A2*RAvg
- UCLx = Target + A2*RAvg
- RAvg = Average range of subgroup (RAvg)
- LCLr = D3*RAvg
- UCLr = D4*RAvg
- D3, D4, A2, d2 are from McNeese. Reference McNeese, W.H., Klein, R.A.,
(1991) Statistical methods for the process industries, Quality and
Reliability/28, chapter 16 table 1, p. 224.
Mean/Standard deviation
To create the Shewhart control chart of type Mean/Standard deviation:
1. Click Control Chart.
2. Select Shewhart in the Type of Control Chart box
3. Select Mean/StdDev deviation in the Shewhart Type box.
4. Enter the desired sample size in the Sample size field (default 5), and then
click OK.
The resulting plot displays the control charts for the selected vector, with the target and
upper and lower limits displayed.

The plot footer displays: Target(Mxx) or Target(UE) when defined, LCL(x),
UCL(x), S(Mxx), S(SAvg) or S(UE) when defined, SAvg(within, S(Mxx) or S(UE)
when defined), LCL(s) and UCL(s).
User Guide to SIMCA
374
To view the nomenclature and calculations of target and standard deviation, see the
Nomenclature and notation and Target and Standard deviation subsections in
Appendix A: Statistics.
Control limits and standard deviation
Control limits and estimates of the standard deviation are computed as follows:
- S(SAvg) = SAvg/c4, estimate of the process standard deviation.
- LCLx = Target - A3*SAvg
- UCLx = Target + A3*SAvg
- LCLs = B3*SAvg
- UCLs = B4*SAvg
- SAvg(within, M1 or S(UE)) = average of the standard deviations of
subgroups, or, when the standard deviation is user entered, c4 multiplied by
S(UE).
- B3, B4, A3, c4 are from McNeese. Reference McNeese, W.H., Klein, R.A.,
(1991) Statistical methods for the process industries, Quality and
Reliability/28, chapter 19 table 1, p. 272.
EWMA Exponentially Weighted Moving Average
The EWMA, Exponentially Weighted Moving Average, control chart can be displayed
both with sample size = 1 and sample size > 1.
Open the EWMA control chart by clicking Control Chart on the Plot/List or Predict
tabs, selecting EWMA in the Type of Control Chart box.

EWMA with sample size = 1
To create the EWMA control chart:
1. Click Control Chart.
2. Select EWMA in the Type of Control Chart box.
3. Leave the default value = 1 in the Sample Size field, and then click OK.
The resulting plot displays the selected vector, target and upper and lower limits in the
EWMA control chart.
Plot/List
375
When plotting t or tPS, a plot of DModX or DModXPS is displayed beneath, with its
critical distance. Observations outside the critical distance indicate process behavior
different from the model.

The plot footer displays: S(Mxx), S(UE) when defined, Lambda (estimated or user
defined), Target(Mxx) or Target(UE) when defined, S(EWMA), UCL(EWMA),
LCL(EWMA), 1 R2X(cum)[last component]and Mxx D-Crit[last component].
To view the nomenclature and calculations of target and standard deviation, or
calculations of limits, see the Nomenclature and notation, Target and Standard
deviation, respective Control limits and standard deviation subsections in
Appendix A: Statistics
EWMA with subgroups
To create the EWMA control chart with subgroups:
1. Click Control Chart.
2. Select EWMA in the Type of Control Chart box.
3. Enter a number > 1 in the Sample Size field and then click OK.
The resulting plot displays the selected vector, target and upper and lower limits in the
EWMA control chart.
With sample size > 1 the EWMA is calculated as:
yhat
t+1
= yhat
t
+ (Y
t
yhat
t
)
where t is time and yhat is predicted y.
User Guide to SIMCA
376

The plot footer displays: S(Mxx), S(UE) when defined, Lambda (estimated or user
defined), Target(Mxx) or Target(UE) when defined, S(EWMA), UCL(EWMA),
LCL(EWMA) and SAvg(between, Mxx or S(UE) if entered).
To view the nomenclature and calculations of target and standard deviation, or
calculations of limits, see the Nomenclature and notation, Target and Standard
deviation, respective Control limits and standard deviation subsections in
Appendix A: Statistics.
Control limits and standard deviation
Control limits and estimates of the standard deviation are computed as follows:
S(EWMA) = MSSD, Mean Square Successive Difference = SAvg * (/(2-))
1/2

The upper and lower control limits are computed as:
UCL(EWMA) = Target + 3 * S(EWMA)
LCL(EWMA) = Target 3 * S(EWMA)
where is estimated (when not user entered) to minimize the error sum of squares.
Plot/List
377
CUSUM
To create the CUSUM control chart:
1. Click Control Chart.
2. Select CUSUM in the Type of Control Chart box.
3. Enter the desired number of groups in the Sample Size field, and then click
OK.
The resulting plot displays the target, CUSUM lines and limits for the selected vector.
The CUSUM, CUmulative SUM, control chart can be displayed both with sample size
= 1 and sample size > 1.


The plot footer displays: S(Mxx), Target(Mxx) or Target(UE) when defined, S(UE)
when defined, Action Limit (H) and Dead Band (K).
To view the nomenclature, calculations of target and standard deviation, or calculations
of limits, see the Nomenclature and notation and Target and Standard deviation
topics subsections in Appendix A: Statistics, respective the Control limits section
later.
User Guide to SIMCA
378
CUSUM plot content
The series displayed in the plot are:
- High CUSUM = Cumulative sum on the high side difference used to detect
a deviation from the target on the high side. High CUSUM is set to zero
when negative.
- Low CUSUM = Cumulative sum on the low side difference used to detect a
deviation from the target on the low side. Low CUSUM is set to zero when
positive.
- Dev from Target, deviation from target = average of subgroup target.
Reference McNeese, W.H., Klein, R.A., (1991) Statistical methods for the process
industries, Quality and Reliability/28, chapter 21.
Note: The plot displays the Low CUSUM with reversed sign (- Low CUSUM)
so that all values are displayed as negatives.
Control limits
The control limits are computed as follows:
- Action Limit (H) = 4.5 standard deviation, using S(Mxx) when sample size
= 1 and SAvg when sample size > 1.
- Dead Band (K) (or allowable slack) = standard deviation/2, using S(Mxx)
when sample size = 1 and SAvg when sample size > 1.
EWMA/Shewhart
Click Control Chart, select Shewhart/EWMA in Type of Control Chart, and then
click OK displays the Shewhart and EWMA lines, target and limits, all in the same
plot.

Note that the EWMA/Shewhart control chart is only available with sample size = 1).
Consequently the Shewhart displayed is the Individual.
Lambda, Target, and Standard Dev. can be User Entered or Estimated. See the
Shewhart and EWMA sections earlier for more.
Plot/List
379

The plot footer displays: S(Mxx), S(UE) when defined, Target(Mxx) or Target(UE)
when defined, S(EWMA), UCL(EWMA), LCL(EWMA) and Lambda (estimated or
user defined). See the Shewhart and EWMA sections earlier for more.
Response Contour
The Response Contour displays the response surface contour plot for the selected y-
variable with the two selected x-variables on the axes, for a PLS model.
To create a Response Contour plot:
1. On the Plot/List tab, in the Custom Plots group, click Response Contour.
2. Select the model in the Data box and the number of components to use in the
Comp box.
3. Select the y-variable to display in the Y-Variable box.
4. Select the x-variables to display on the x-axes in the 1
st
axis and 2
nd
axis
boxes. The range of the selected variables is displayed using Low and High
settings fields. These are the values used to limit the surface. Change them as
warranted.
5. All x-variables not selected to be displayed on the axes, are held constant and
listed in the Constant variables list, by default set at their average, Center.
Change these settings as warranted by typing new values in the respective
fields.
6. Click OK or click the 3D Plot Options tab (described next). After clicking
OK, the values of the constant factors are displayed near the plot. Changing
these values updates the plot instantly.
User Guide to SIMCA
380

Response Contour plot options
Clicking the 3D Plot Options tab enables changing the following options:
1. Resolution 16, 32, 64, or 128. Resolution 32 is the default.
2. Show contour level labels - clear the check box to hide the labels in the plot.

See also the Contour subsection in the Layout tab section in the Plot and list
contextual tabs chapter.
Plot/List
381
Response Surface Plot
The Response Surface plot displays the response surface contour plot in a 3D grid, for
the selected y-variable with the two selected x-variables on the x and y-axes and the y-
variable on the z-axis. The plot is only available for PLS, OPLS and O2PLS models.
To create the Response Surface plot, on the Plot/List tab, in the Custom Plots group,
click Response Surface. The dialog that opens is the same as the one for the Response
Contour, see the Response Contour section preceding this section.

Wavelet Structure
In the Wavelet Structure plot the selected vector is decomposed using DWT into its
approximations and details at every scale. By default, the details and approximations
for the first four lowest scales, scales 1 to 4 (the first 4 highest frequencies bands), are
reconstructed and plotted using the same time axis. The original signal is A0. This is
very useful in order to understand the structure of a signal.
Note: In SIMCA the scales or levels are the inverse of the frequencies. Hence,
scale 1 (D1) corresponds to the highest frequency band in the signal and the
highest scale corresponds to the lowest frequency in the signal. The mean (DC
component) is not included.
User Guide to SIMCA
382

The bar chart plot displays the percent of the total Sum of Squares of the signal (not
including the DC component) contained in the details coefficients at every scale. Plots
A1 to A4, and D1 to D4 are the reconstructed approximations and details coefficients
at the scale 1 to 4. A0 is the original signal.
Displaying Wavelet Structure plot
To display the Wavelet Structure plot:
1. Click Wavelet Structure.
2. Select the vector to display.
3. Click the Wavelet Options tab and select detrending, wavelet family, and
wavelet order. For more, see the Wavelet Options subsection in the Data
chapter.
4. Click the Wavelet Details tab and select the levels to display. For more, see
the Wavelet compression/denoising By detail level subsection in the Data
chapter.
5. Click OK.
Wavelet Power Spectrum
The wavelet power spectrum plot is a 3-D plot displaying the scaled power density (the
normalized squared wavelet coefficients) of the signal as a function of both time and
frequency. Note that the frequencies are in logarithmic band passes (displayed on the
plot as multiple of 2) up to the Nyquist frequency (the highest frequency in the signal).
The smallest even scale (i.e. 2) on the plot represents the highest frequency band, that
is frequencies in the signal from the Nyquist up to the Nyquist frequency. The
highest even scale in the plot is the single lowest frequency in the signal.
The DC component (average) when present is not included.
Plot/List
383
Displaying Wavelet Power Spectrum plot
To display the Wavelet Power Spectrum:
1. Click Wavelet Power Spectrum.
2. Select the vector to display.
3. Click the Wavelet Options tab and select detrending, wavelet family, and
wavelet order. For more, see the Wavelet Options subsection in the Data
chapter.
4. Click OK.
Wavelet Power Spectrum example
The following plots display a signal, which is 20-Hertz in the first half, and 100-Hertz
in the second half. The sampling frequency of the signal is 1024- Hertz (i.e. 1024
points a second).
In the Wavelet Power Spectrum, the 20-Hertz signal is displayed as continuous peaks
across the first half of the time axis, on scale 10, i.e. frequency band between 16 and 32
Hertz. The 100-Hertz signal occurs across the second half of the time axis, with peaks
centered on scale 6, i.e. frequency band between 64 and 128-Hertz.

User Guide to SIMCA
384
Step Response Plot
The Step Response Plot v
k
is the sum or integral of the Impulse Response function v
k

of the system. These are the response of the system at times t and k >0 to a unit pulse
input at time t =0.
Hence
Y
t
= v
k
X
t-k

k=0 to is the deviation from steady state at time t.
Creating a finite impulse response model
When clicking Step Response Plot, a wizard opens with 2-6 pages depending on the
selections in the wizard. The pages up to and including the page specifying the lag
structure define the FIR model.
Selecting to create a new model
On the first page of the Step Response Plot wizard, select to create a new model by
clicking Create a new Finite Impulse Response model. Then click Next.

Selecting y-variable and excluding variables or
observations
On the 2
nd
page of the Step Response Plot wizard:
- Select the y-variable in the Variables list, and click the Y-button.
- Exclude undesired variables and observations by selecting them and clicking
the respective Exclude-buttons.
- Optionally, select the Change transformation check box to transform
variables. For more, see the Applying transformations subsection in the
Workset section in the Home chapter.
- Optionally select the Change scaling check box to change the scaling of
variables. Then click Next, and customize the scaling using the None,
Pareto, Unit Variance, and Center-buttons.
Plot/List
385
When done click Next.

Specifying lag structure and fitting FIR model
To specify the new models lag structure:
1. The default Maximum number of lags is 45 or N-5 whichever is smaller.
Specify a new maximum number of lags by typing a new value smaller than
N-5.
2. The default Lag Increment is 5 or Maximum number of lags/3 whichever
is smaller. Specify a new lag increment by typing a new value smaller than
Maximum number of lags/3.
3. Click Next and SIMCA fits an FIR model with the selected y-variable and all
x-variables with their lags as selected.

User Guide to SIMCA
386
Displaying Step Response Plot
On the last page of the Step Response Plot wizard:
1. Select the x-variable to display in the Available X-variables list.
2. In the Coefficients as a function of lag plot, mark the integration interval to
display in the Step Response Plot. Leaving the plot as it is leads to
displaying the entire integration.
Note that the Plot Preview window displays the Step Response Plot and is updated
when switching x-variables and integration intervals.
When the FIR model is fitted the wizard displays the Impulse Response function of the
selected x-variable.

Click Finish when done to display the final Step Response Plot. In this plot the
integration always starts at 0.

Plot/List
387
Using an existing finite impulse response model
To use an existing Finite Impulse-Response model to display another Step Response
Plot:
1. Click Step Response Plot.
2. Click Use an existing Finite Impulse Response model and select one of the
models listed in the Available Finite Impulse-Response list.
3. Click Next to open the last page of the wizard. For more about this page, see
the Displaying Step Response Plot subsection previously in this section.


389
View
Introduction
This chapter describes all commands on the View tab.
The following is available:
- The Show/Hide group including all dockable windows, the Model Window
and the Full Screen command.
- The Add group including Add to Favorites and Add to Report.
- The Window group including Cascade, Tile and Close of windows.
- Optionally the Skins group if there are skins installed.

Show/Hide
In the Show/Hide group on the View tab you can select to show or hide the following:
Audit Trail, Analysis Advisor, Favorites, Item Information, Marked Items, Model
Window, Observations, Project Window, Quick Info, Variables, and the Status
Bar. Clicking Full Screen minimizes the ribbon as when Minimizing the ribbon.

All check boxes represent dockable windows except for the Model Window which is a
regular window, and the Status Bar which is always positioned at the bottom of
SIMCA and displays the plot coordinates of the cursor or the first part of the current
tooltip.
After selecting one of the dockable windows, you can move the window and dock it as
desired.
To auto hide the window when not using it click the Auto Hide button . The button
changes to and the window will slide away when not used and slide up on top of
plots when clicked.
The features in the Show/Hide group are described in this section.
User Guide to SIMCA
390
Audit Trail
The Audit Trail was implemented in SIMCA, in compliance with rules for keeping
electronic records (21 CFR part 11).
When turned on, the Audit Trail window logs all events in a session. A session starts
with the creation or opening of a project and ends when the project is saved. When a
project is reopened the current event logging is appended to the existing audit trail.
In addition to logging events, SIMCA logs information about the user and date and
time of the events.

Turning on logging to the Audit Trail
To turn on the logging to the Audit Trail:
- Before selecting any datasets in the SIMCA import, click File | Options and
select Yes in the Log events to the Audit trail box. The audit trail will then
be turned on for this specific project and by default for all projects created
thereafter.
- For projects already open in SIMCA, click File | Project Options, click the
Audit Trail tab, and then select the Log events to the Audit Trail check
box. The audit trail will then log the events in this project that follow after it
was turned on. This has no effect on projects created after.
The administrator can lock the audit trailed turned on. For administration of the Audit
trail, see the Administration of the audit trail subsection in the Project Options
section in the File chapter.
The audit trail logs the actions listed in the Logged in the Audit Trail subsection next.
Logged in the Audit Trail
The following actions are logged by the Audit Trail:
Import and changes of the dataset during import/dataset creation:
- name and path of imported file(s).
- changes to the datasets done in SIMCA import, i.e. values changed, rows
excluded etc.
- settings for creating batch level dataset.
View
391
Creation/changing of a workset:
- shows total number of X-variables.
- number of expansions.
- number of lags.
- number of Y-variables.
- number of observations.
- model type.
- number of excluded observations and variables.
- number of transformed variables.
- the datasets selected.
Changes of the model, such as:
- change in number of components.
- hierarchical model type.
- options.
- deletion of a model.
Creation/changes of datasets:
- logs the old and new values when changing values.
- logs the old and new value when changing observation or variable IDs.
- deletion of observations and variables.
- addition of observations and variables.
- sorting.
- splitting.
- merging.
- transposing.
- deletion of datasets.
- addition of variables using Generate variables on the Data tab.
- logs when there is a change in predictionset.
The data source name when the data comes from a database.
Save and Save As (including the name of the new project).
Registers when a digital signature in the Audit trail is incorrect.
Clearing of the Audit trail (File | Project Options, Clear Audit trail).
Turning on and off the Audit Trail.
Changes in the Audit Trail extended user information.
User Guide to SIMCA
392
Analysis Advisor
The Analysis advisor explains all plots available on the Home tab. The Analysis
Advisor window by default opens on the right side of the screen.

Favorites
Open the Favorites window by adding a plot or list to it, or by selecting the Favorites
check box in the Show/Hide group on the View tab.

The Favorites window contains plots. Clicking a plot opens the specified plot for the
selected model in the current project.
To move items or folders, grab and drag them.
View
393
The following commands are available from the Favorites shortcut menu:
- Open when a single item is marked.
- Open All Items in Folder when a folder is marked.
- Treat Folder as Item when a folder is marked. Toggles treating the folder as
an item on or off.
- Rename folder or item.
- Delete folder or item.
- New Folder.
- Export the Favorites configuration to file.
- Import the Favorites configuration from file.

Organizing in folders in Favorites
Plots can be organized in folders in the Favorites window. There is one global folder
available by default, Favorite Plots and Lists, and one local folder, Project
Favorites.
The default folders can be customized, and new global or local folders can be created.
Favorites positioned in Project Favorites are saved with the current project and do not
remain when another project is opened.
Opening plots, items, and executing folders in Favorites
Each item in the Favorites window can be opened by double-clicking it or by marking,
right-clicking, and clicking Open.
All items or commands in a folder can be opened, and the windows tiled, by marking
the folder, right-clicking it and clicking:
- Open All Items in Folder.
- Treat Folder as Item and then double-clicking it.
Renaming in Favorites
Rename folders or items by:
- Right-clicking the item and selecting Rename.
Or
- Marking the item and pressing F2.
User Guide to SIMCA
394
Deleting from Favorites
To delete a folder or an item, mark the item and:
- Right-click and select Delete.
- Press DELETE on your keyboard.
Creating folders in Favorites
It is convenient to group plots in folders, and automatically open all plots in the folder
in sequence when often opening the same plots.
To create a folder:
1. Right-click in the Favorites window.
2. Click New Folder and type a name for the folder.

Exporting and importing Favorites configuration
The Favorites window configuration can be saved as an .xml-file.
To save the current favorites configuration to file:
1. Right-click the Favorites window.
2. Click Export.
3. Enter the name and location in the Save As dialog; click Save.
To import favorites from .xml-file:
1. Right-click the Favorites window.
2. Click Import.
3. Browse for the file in the Open dialog; click Open.
Note: Importing a favorites file overwrites the current file.
Adding plots and lists to Favorites
To add a plot or list to Favorites use one of the following methods:
1. Right-click the plot or list, and then click Add to Favorites.
2. With the plot or list active, press CTRL+D.
3. With the plot or list active, on the View tab, in the Add group, click Add to
Favorites.
View
395
Item Information
The Item Information window displays all the information about the items that are
marked in a plot.

SMILES
If you have SMILES code as a secondary Observation ID, and have the SMILES plug-
in DLL, the Item Information window displays the molecule.
When marking several observations holding down the CTRL-button, the Item
Information window only displays the chemical structure for the last marked
observation.
To see the structure of several observations, all points need to be marked in one go.
When the points are not adjacent:
1. Mark the points you want to see structure for.
2. Right-click, select Create Plot or Create List.
3. Mark all in observations in one scope (not holding CTRL-button) and the
chemical structure is displayed for all observations in Item Information.
For more information about SMILES, contact your Umetrics sales office.

User Guide to SIMCA
396
Marked Items
The Marked Items window displays all identifiers of the currently and secondary
marked items.

The Marked Items window can be used for many purposes. For instance:
- To display the items, observations or variables, marked in a plot to visualize
which items are actually marked. In plots with many items it can be difficult
to see precisely which those items are.
- To visualize the marked points in the Marked Items window when marking
points to create a contribution plot.
- To unmark (deselect) points, mark the points in the Marked Items window,
right-click the marking, and click Unmark Items.
- To view the groups in the dendrogram plot, opened by clicking HCA in the
Clustering group on the Analyze tab.
Marked Items window features
There are a number of features available in the Marked Items window shortcut menu.
With observations (left below) in Marked Items, the shortcut menu holds the
commands: Unmark Items, Include, Exclude, Set Class, Select All, Clear All
Marked, and Observation Label.

With variables (right above) in Marked Items, the shortcut menu holds the
commands: Unmark Items, X, Y, Exclude, Select All, Clear All Marked, and
Variable Label.
View
397
Clicking Include, Exclude, Set Class, X, Y, or Exclude creates a new unfitted model
unless there already is an unfitted model. If there already is an unfitted model the
selected command is done for that model.
When a new unfitted model is created, that model is created as a copy (as New As in
the Workset group) of the model used when marking the items displayed in the
Marked Items window.
Select Result
Unmark Items Unmarks the marked items in the Marked Items window
both in the window and in all open plots and lists.
Include Creates a new model including only the marked items.
Exclude Excludes the marked items.
Set Class Select No Class or a class. Organizes in classes.
X Sets the marked variables as X.
Y Sets the marked variables as Y.
Select All Marks all items in the Marked Items window.
Clear All Marked Clears all marking in the window and in all open plots and
lists.
Observation
Label and
Variable Label
The Marked Items window lists all identifiers. Select to
remove labels under Observation Label or Variable
Label.
Creating contribution plots using Marked Items window and
tab
The Marked Items window is useful when creating contribution plots, and especially
group contribution plots.
When creating a group contribution plot, the observations in the group marked first are
displayed under Secondary Marked Values and the group marked last under Marked
Values.
When marking observations the Marked Items tab becomes available and active.
To create a contribution plot, click the relevant button in the Drill down group.
For more about creating contribution plots by marking in plots, see the Creating plots
from plots section in the Marked Items tab section in the Plot and list contextual
tabs chapter.
User Guide to SIMCA
398
Model window
The model window includes an overview of the model, the Workset and Options
buttons, and displays a summary of the fit of the model with results for each
component.

The model window holds the following information about the model:
Item Explanation
Title Title of model.
Type The model type used when fitting.
Observations
(N)
Number of observations.
Variables (K) Total number of variables
X Number of variables as X (including expanded and lagged terms).
Y Number of variables as Y.
Lagged Number of lagged variables.
Expanded Number of expansions.
Components The list of components and their statistics and properties. For
more see the Model summary of fit subsection later in this
section.
Workset button
Clicking the Workset-button displays the workset associated with the model. Note
that if you click the Workset-button, edit the workset, and click OK, the model will be
replaced by a new unfitted model. For more about the workset, see the Workset
section in the Home chapter.
Options button
Clicking the Options button displays the Model Options dialog. For details, see the
Model Options section in the Workset section in the Home chapter.
View
399
Model summary of fit
The model window consists of a summary line for each component A of the
model, starting with component 0 representing centering.
The table describes all columns available for each component.
Column
header
Description
A The component number.
R2X Fraction of Sum of Squares (SS) of the entire X explained by the
current component.
R2X (cum) Cumulative SS of the entire X explained by all extracted
components.
Eigenvalue Eigenvalue of the X matrix, R2X * min(K,N).
R2Y Fraction of Sum of Squares of all y-variables explained by this
component.
R2Y (cum) The cumulative SS of all the y-variables explained by the
extracted components.
Q2 The fraction of the total variation of X (PC) and Y
(PLS/OPLS/O2PLS) that can be predicted by the current
component.
Limit The cross validation threshold for that component. When Q2 >
Limit the component is significant.
Q2 (cum) The cumulative Q2 for all the x-variables (PC) and y-variables
(PLS/OPLS/O2PLS) for the extracted components.
Significance Significance of the component according to cross validation rules:
R1, R2, U, R5, NS, N3, N4. For details about the cross validation
rules, see the Cross validation section in Appendix A:
Statistics.
Iterations Number of iterations till convergence.
Model window for OPLS and O2PLS models
The Model Window corresponding to OPLS and O2PLS displays summary statistics
related to the model. Components that capture variation found in both X and Y are
denoted Predictive. Components that capture variation only found in X are denoted
Orthogonal in X(OPLS). Components that capture variation only found in Y are
denoted Orthogonal in Y(OPLS).
The OPLS model is based on the OPLS algorithm. The O2PLS model is based on the
similar O2PLS algorithm, but in addition there is a PCA step. PCA is used after
convergence of the O2PLS algorithm, to exhaust the E and F residual matrices from all
remaining systematic variation. This yields the additional Orthogonal in X(PCA) and
Orthogonal in Y(PCA) components.
The method underlying a certain Orthogonal component is indicated in the name of the
component in the Model Window.
User Guide to SIMCA
400
Note: The model window for OPLS models displays, if existing, three types of
components: Predictive, Orthogonal in X, Orthogonal in Y. Additionally, The
model window for O2PLS displays, if existing, Orthogonal in X and
Orthogonal in Y components estimated by PCA.
Figure 2. The Model Window for an O2PLS model with 7 predictive, 3 Orthogonal in
X and 6 Orthogonal in Y components (a 7+3(1+2) +6(1+5) O2PLS model).

The Model Window displays:
Section Description Component types
Model Summarizes the model,
showing the cumulative R2X,
R2, Q2, and R2Y.

Predictive The Predictive section where
the first row summarizes the
predictive components in the
model followed by a listing of
each predictive component.
The predictive loading vectors are
p for the X-block and q for the Y-
block. The predictive score vectors
are t for the X-block and u for the
Y-block. In the figure above there
are 7 components for p, q, t and u.
Orthogonal
in X (OPLS)
The Orthogonal in X(OPLS)
section where the first row
summarizes the orthogonal
components in the X model
followed by a listing of each
orthogonal in X component.
The orthogonal in X
components show the
variation in X that is
uncorrelated to Y.
The orthogonal in X(OPLS) loading
vectors are po for the X-block and
so for the Y-block. The orthogonal
in X (OPLS) score vector is to for
the X-block.
In the figure above there is one
orthogonal in X(OPLS) component.
This means that the po[1], so[1],
and to[1] vectors are the
orthogonal in X(OPLS) vectors.
View
401
Section Description Component types
Orthogonal
in X (PCA)
The Orthogonal in X(PCA)
sections where the first row
summarizes the orthogonal
components in the X model
followed by a listing of each
orthogonal in X component.
The orthogonal in X
components show the
variation in X that is
uncorrelated to Y
The orthogonal in X(PCA) loading
vectors are po for the X-block and
so for the Y-block. The orthogonal
in X (PCA) score vector is to.
The orthogonal in X(PCA)
components are extracted after the
orthogonal in X(OPLS)
components, and their relational
order is indicated through the
shared nomenclature.
In the figure above there are two
orthogonal in X(PCA) components.
This means that the po[2], po[3],
so[2], so[3], to[2] and to[3] vectors
are the orthogonal in X(PCA)
vectors
Orthogonal
in Y (OPLS)
The Orthogonal in Y (OPLS)
sections where the first row
summarizes the orthogonal
components in the Y model
followed by a listing of each
orthogonal in Y component.
The orthogonal in Y
components show the
variation in Y that is
uncorrelated to X.
The orthogonal in Y(OPLS) loading
vectors are r for the X-block and qo
for the Y-block. The orthogonal in Y
(OPLS) score vector is uo for the Y
block.
In the figure above there is one
orthogonal in Y(OPLS) component.
This means that the r[1], qo[1], and
uo[1] vectors are the orthogonal in
Y(OPLS) vectors.
Orthogonal
in Y (PCA)
The Orthogonal in Y (PCA)
sections where the first row
summarizes the orthogonal
components in the Y model
followed by a listing of each
orthogonal in Y component.
The orthogonal in Y
components show the
variation in Y that is
uncorrelated to X
The orthogonal in Y(PCA) loading
vectors are r for the X-block and qo
for the Y-block. The orthogonal in Y
(PCA) score vector is uo for the Y
block.
The orthogonal in Y(PCA)
components are extracted after the
orthogonal in Y(OPLS)
components and their relational
order is indicated through the
shared nomenclature.
In the figure above there are five
orthogonal in Y(PCA) components.
This means that the r[2]-r[6], qo[2]-
qo[6], and uo[2]-uo[6] vectors are
the orthogonal in Y(PCA) vectors.
The table columns are:
- Component - Component index.
- R2X - Fraction of X variation modeled in that component, using the X
model.
- R2X(cum) - Cumulative R2X up to the specified component.
- Eigenvalue - The minimum number of observations (N) and X-variables
multiplied by R2X, that is, min(N,K)*R2X.
User Guide to SIMCA
402
- R2 - Fraction of Y variation modeled in that component, using the X model.
- R2(cum) - Cumulative R2 up to the specified component.
- Q2 - Fraction of Y variation predicted by the X model in that component,
according to cross-validation.
- Q2(cum) - Cumulative Q2 up to the specified component.
- R2Y - Fraction of the Y variation modeled in that component, using the Y
model.
- R2Y(cum) - Cumulative R2Y up to the specified component
- EigenvalueY - The minimum number of observations (N) and Y-variables
multiplied by R2Y, that is, min(N,M)*R2Y.
- Significance Significance level of the model component.
For more, see the OPLS/O2PLS - Orthogonal PLS modeling section in Appendix
A: Statistics.
Observations window
The Observations window displays the observations included in the workset of the
active model.
Right-clicking the window opens the shortcut menu holding the following commands:
Include, Exclude, Set Class, Select All, and Observation Label. These commands
work as described in the Marked Items section previously in this chapter.

Project window
The project window displays a list of all models with their respective model
information.
The model selected in the project window is the active model, displayed in the window
caption (frame) both when displayed and when minimized. All plots created are
created for the active model.
View
403
The project window is by default docked and can then only be minimized, not closed.
To have the project window slide away after showing it, click the Auto Hide button so
that it turns. See the Show/Hide subsection for details.
To have the project window as a regular window, see the Customizing the Project
window subsection.

The list of models displays the following information about each model:
Column Description
No. Workset number. Workset numbers are assigned sequentially.
When the model is a batch model the workset number also holds
the abbreviation BM for Batch Model. With class models, the
workset number also holds the abbreviation CM for Class Model.
Model Automatic model name. The models are named Mxx with xx
being a sequential number starting at 1
Type Fit method used to fit the model.
A Number of components.
R2X Sum of Squares of all the x-variables explained by the extracted
components.
R2Y For PLS, Sum of Squares of all the y-variables explained by the
extracted components.
Q2(Cum) Cumulative cross validated R2.
SD(Y) Standard deviation of Y (not displayed by default).
Date Date when the model was fitted.
Time Time when the model was fitted (not displayed by default).
Title Title of the model entered by the user. By default all models have
the title 'Untitled'.
Hierarchical Displays B or T when the model is a Base or Top hierarchical
model. A model using the scores or residuals of other models as
variables is a hierarchical top model. Hierarchical models are
described in detail in the Analyze chapter.

User Guide to SIMCA
404
Customizing the Project window
To select which columns to display in the Project Window, click Customize from the
shortcut menu.
In the dialog that opens, select and clear as desired.

Project Window shortcut menu
From the shortcut menu of the Project Window the following commands are
available:
- Open - Opens the model window of the active model.
- Edit Model x - Opens the workset for the model, see the Editing the
workset subsection in the Workset section in the Home chapter.
- New as Model x - Opens a copy of the workset for the model, see the New
workset as model subsection in the Workset section in the Home chapter.
- Change Model Title - Opens the Model Title dialog, see the Changing the
model title subsection later in this chapter.
- Delete - Deletes the active model, see the Deleting the workset subsection
in the Workset section in the Home chapter.
- Hierarchical Base Model, Non Hierarchical Base Model - Specifies the
model as hierarchical/non hierarchical base model, see the Hierarchical
models subsection in the Data chapter.
- Generate Report - Opens the report generator, see the Appending to,
inserting in, or replacing existing report subsection in the Generate
Report section in the File chapter.
- PLS-Tree Plot - Opens the PLS-Tree dendrogram, see the PLS-Tree
resulting dendrogram subsection in the Analyze chapter.
- Add to Report - Adds the project window to the report, see the Add to
Report subsection in the Generate Report section in the File chapter.
- Customize - Opens the Customize Project Window dialog, see the
Customizing the Project window subsection later in this section.
- Model Options - Opens the Model Options of the active model, see the
Model Options subsection in the Workset section in the Home chapter.

View
405
Changing the model title
The model title is by default 'Untitled'. To change the model title:
1. Use one of the following methods:
- With the Model Window open, type in the Title field.
- With the Project Window open, right-click the model, and click
Change Model Title.
2. Enter the new model title and click OK.
Quick Info
The Quick Info window displays overview information about marked items in an open
plot, list, or spreadsheet. When displaying the Quick Info of variables after marking in
a dataset spreadsheet, trimming and Winsorizing is also available.
A separate Quick Info window is available in the workset dialog, tab Spreadsheet.
Trim-Winsorizing in the workset applies to the model only, not the dataset. For more,
see the Spreadsheet in the Workset dialog subsection in the Workset section in the
Home chapter.
To open the Quick Info dockable window:
- On the View tab, in the Show/Hide group, select the Quick Info check box
OR
- Right-click the dataset spreadsheet and click Quick Info.

User Guide to SIMCA
406
Quick Info window description
The Quick Info window has the following content:
- Caption bar with the current item information including the type of item
(variable/observation etc), Auto Hide, and Close.
- Statistics section holding the statistics and optionally variable recipe when
selected for display in Options. Default is to display N (number of items in
the series), % Mis. Val. (percent missing values), Mean, and Std. dev.
(standard deviation).
- Selection of plots according to selection in Options. Default is to display the
Frequency Histogram and Time Series plots.
- Selection Bar if selected in Options. Default is to not display the Selection
Bar.
- Options button.
- Trim-Winz Var and Trim-Winz All-buttons when the Quick Info is for
variables.
The quick info plots and the buttons Options, Trim-Winz Var, and Trim-Winz All
are described in the subsections that follow.
Quick info plots and interactive delete or replace
Deleting an observation or variable, or changing the value of a selected cell, can be
done interactively using the Quick Info window Time Series or Spectrum plot in the
following manner:
1. Mark points in the Time Series or Spectrum plot. The Remove or Replace
the x Selected Observations/Variables dialog opens:

2. Select
a. Remove the variable/observation from the dataset to delete the
marked variables or observations OR
b. Replace the value in the cell in the dataset and type the new
value in the New value cell, to replace.
3. Click OK to take action.
Note: The Remove or Replace dialog automatically opens when marking
points in the Quick Info Time Series or Spectrum plots when the Trim-Winz
Var dialog is closed.
Quick Info Options button
The selected settings in the Quick Info Options dialog define what is displayed in the
Quick Info window.
View
407
Clicking the Options button opens the options dialog with the tabs Quick Info
Options, Power Spectrum Options, and Correlation Options. The content of these
tabs is described in the subsections that follow.
Quick Info Options page
To open the Quick Info Options dialog, click the Options button.

Statistics section
The first section under Display lists the statistics that can be displayed in the Statistics
section of the Quick Info window and an option named Variable recipes.
The available statistics are: N, Missing value (%), Min, Max, Min/Max, Mean,
Median, Standard deviation, Std. dev/Mean, Skewness, and Kurtosis.
The Variable recipe describes how a new generated variable was created. For
example, after creating VAR16 as the product of variable v5 multiplied with the ratio
of variable v10 and variable v1, clicking that variable displays the formula in the
statistics section.

Selecting Quick Info window plots
Select to display any or all of the 4 following plots:
1. Frequency histogram (default selected).
2. Spectrum/Time series (default selected).
3. Auto correlation.
4. Power spectrum.
User Guide to SIMCA
408
Displaying the selection bar
The Selection bar is a Variable selection bar with a scrolling tool. Moving this tool
allows moving in the dataset by emulating cursor movement. Above the scroll bar, the
current variable is displayed in the Variable selection box. Use the Variable selection
box or click the arrow-button to switch variables.

Selecting items to include
Under Use only on the right side of the window is a list of observations (for Quick
Info of variables) or variables (for Quick Info of observations). Use this function to
select which observations or variables to include. By default all the observations or
variables are included.
A Find feature is available to facilitate this task. Observations or variables may be
selected by entering characters in the Find field. Wild card symbols '?', and '*' are
allowed in specifying observations or variables identifiers. For example ?LH* selects
observations or variables with names such as SLH2 or QLHSW, etc. Click the
Complement button to display the complement of the selected variables or
observations.
Note: The Find utility in SIMCA is cASe inSEnitiVE.
To display another or more identifiers, right-click the list and select it.
Selecting phases and batches for batch data
With batch data, select to display the quick info window for all phases and batches or
for a selection of phases and batches.

View
409
Select the phase and batches from the boxes positioned bottom right. Default is to use
All Phases and All Batches. All statistics and plots will refer to the selected phases
and batches.
Power Spectrum Options page
The Power Spectrum Density (PSD) is the representation of the sequence x (t) in the
frequency domain. The power spectrum is the square of the amplitude of the Fourier
component at each frequency.
The frequency domain representation is particularly helpful to detect irregular cycles
and pseudo periodic behavior, i.e. tendency towards cyclic movements centered on a
particular frequency.
For more, see the Power Spectrum Density section in the Appendix B:
Preprocessing chapter.
Note: With a dataset that has been wavelet transformed and compressed
variables wise, using Data | Spectral Filters, the PSD observation wise refers
to the reconstructed observations, when the Reconstruct wavelets check box is
selected.

The Power Spectrum Options tab includes the following options.
Detrending
The default is to detrend by removing the Mean. There are two more options:
- Removing the best linear fit by selecting Linear or
- Removing nothing by selecting None.
Scaling
Asymptotically Unbiased is the default scaling. Selecting Peaks estimates the heights
of the original peaks.
Selecting window type
The default Window type is Hanning.
To select another window type, click the Window type box and select Welsch or
Bartlett.
Selecting window length
The default Window length is the smallest of 256 or the length of X (t), padded, to
the nearest length 2
n
, where n= integer, although always larger than or equal to 16.
Select another length from the Window length box.
User Guide to SIMCA
410
Overlapping segments
The default is to not overlap the segments.
To overlap, select the Overlapping segments check box.
Scaling of the Y axis
The scale of the y-axis is default in Decibel.
To display the y-axis non transformed, clear the Amplitude in decibel check box.
Selecting sampling frequency
The output of the PSD is a vector of length = (window length)/2 +1. This is the scale of
x if you have not entered a sampling frequency.
The sampling frequency is used as a scaling multiplier to properly scale the frequency
axis, with the highest frequency being the Nyquist frequency = Fs/2. It has no effect on
the PSD.
By default SIMCA assumes a sampling frequency of 1 and Fs/2=0.5 is displayed as the
highest frequency.
Correlation Options page
The auto correlation is a measure of dependence of adjacent observations and
characterizes a time series or observation profile, in the time domain.
Note: With a dataset that has been wavelet transformed and with the
Reconstruct wavelets check box selected, the auto correlation is displayed in
the reconstructed domain.
For more, see the Auto and cross correlation of variables or observations section in
Appendix B: Preprocessing.

The Correlations Options page includes the following options:
Detrending
The default is to detrend by removing the Mean. There are two more options:
- Removing the best linear fit by selecting Linear or
- Removing nothing by selecting None.
Maximum number of lags
The auto correlation is default computed up to lag L=30 or N/4 whichever is smaller.
To change the number of lags, enter a new number in Max lag.
Selecting plot type
Default is to display the auto correlation plot as a column plot.
To display it as a line plot, select Line in the Plot type box.
View
411
Preprocessing the dataset using Trim-Winsorizing
Trimming and Winsorizing is available for variables only. The Trim-Winz Var and
Trim-Winz All buttons are available when one or more variables are marked.
Trimming is cutting the upper and lower edges of the variables and replacing them by
missing values.
Winsorizing is cutting the upper and lower edges of the variables and replacing the
removed values by new values.
Note: Trimming or Winsorizing the dataset deletes all models that include the
affected variables.
The Trim-Winz Var and Trim-Winz All buttons are positioned bottom right in the
Quick Info window:

To Trim-Winsorize, click:
- Trim-Winz Var to trim or Winsorize the marked variable.
- Trim-Winz All to trim or Winsorize all variables in the dataset or a selected
subset by using the Selected Variables page.
Note: When a number of values are equal, none of them will be changed if not
all of them fall outside the limits.

Note: The details of the performed trim-Winsorizing are registered in the
audit trail when the audit trail is turned on and in the Trimming Overview.
Trim-Winz Var dialog
The Trim-Winz Var dialog displays:
- By section where you can select either the By values option or the By
observation numbers option. The trimming and Winsorizing By values is
applied to values found outside the defined limits. Observations found within
the limits remain untouched. Trimming and Winsorizing By observation
number is applied to the items found inside the defined limits and the items
found outside the limits remain untouched.
- Upper limit section defines the upper limit including limit type, limit,
optional new value, and new value type.
- Lower limit section defines the lower limit including limit type, limit,
optional new value, and new value type.
- Undo trimming button resets the performed trimming or Winsorizing with
the exception when selecting Deleting Obs. Deleting Obs. cannot be reset.
- Apply button performs the defined trimming or Winsorizing. The statistics
are updated and displayed for the changed dataset.
- Cancel button cancels any unapplied trimming and closes the trim-Winz
dialog.
User Guide to SIMCA
412
Note: The Trim-Winz Var dialog remains open after clicking Apply to
facilitate trim-Winsorizing of another variable.
More about the upper and lower limit sections follows.

Defining upper and lower limits By values
The Upper limit and Lower limit sections are identical. To define the upper and lower
limits By values follow the steps that follow:
1. Under Metric select one of the options:
a. Std Dev. to define and view the Cut-off value in robust standard
deviation. The robust standard deviation is computed as the
interquartile divided by 1.075.
b. Value to define and view the Cut-off value as the real value.
2. In the Cut-off value or Percentage fields, enter the value defining the cut
off. When entering a value in the Cut-off value field, the Percentage field is
automatically updated and vice versa. After defining the limit, note that the
number of points affected is displayed in the Number of points field.
3. Select what to do with the marked items by selecting one of the options
available from the Replacement options menu, described in the table, and
then click Apply.

View
413

Replacement
options
New value Result and example
Std. Dev. Number of robust
standard
deviations.
Entering for instance 2 means that the
limit is 2*robust standard deviation
where the robust standard deviation is
calculated as the interquartile/1.075.
To view the actual number that will be
used when replacing:
1. Select Std Dev under Metric.
2. Enter the number of standard
deviations entered in the New
value field, in the Cut-off value
field.
3. Click Value under Metric to
automatically display the standard
deviations value in real values in
the Cut-off value field.
Value The actual value
to replace with.
Enter the value 1 replaces all marked
values with 1.
To replace with missing, leave New
value blank.
Percent Percentage value. Replaces with the value at the entered
percentage.
To view the actual number that will be
used when replacing:
1. Select Value under Metric.
2. Enter the percentage value
entered in the New value field, in
the Percentage field. The Cut-off
value field is then updated to
display the real value.
Last good value No entry. Valid with Time Series and defined as
the last value (in time) of this variable
that was inside the limit.
Delete Obs No entry. Irretrievably deletes the observations
from the dataset.
Note: When switching between displaying the limit in Std Dev. and Value the
Cut-off value field is updated automatically. That is, if the limit is the value
10, clicking Std Dev. automatically displays the corresponding value for
standard deviations.
User Guide to SIMCA
414

Defining upper and lower limits By observation numbers
When selecting By observation numbers the fields available are the upper and lower
limits, the Replacement options box and the New value field.
1. In the Cut-off value fields, enter the observation numbers defining the lower
and upper limits for the observations to trim or Winsorize.
2. Select how to trim or Winsorize the marked items by clicking the arrow and
selecting one of the items. For details, see table in the previous section
Defining upper and lower limits By values.
3. Click Apply.
View
415
Trim-Winsorizing from Quick Info plot By values
When trimming or Winsorizing interactively By values, the area outside the marking
will be affected.
To trim-Winsorize directly from the Quick Info plots Frequency Histogram or Time
Series:
1. Click the Trim-Winz Var button.
2. Select the By values option to mark an area according to the value.
3. In the histogram, vertically mark the area of the plot. Alternatively, in the
time series plot, horizontally mark the area of the plot to keep. The area
becomes white and the area outside the cut off limits becomes yellow. The
Trim-Winz Var dialog is updated according to the marking displaying the
cut-off values, percentages, and the number of marked data points.

4. In the Replacement options boxes select how to replace and if applicable
enter a value in the New value fields.
5. Click Apply.
User Guide to SIMCA
416
Trim-Winsorizing from Quick Info plot By observation numbers
When trimming or Winsorizing interactively By observation numbers, the marked
area will be affected.
To trim-Winsorize directly from the Quick Info plot Time Series:
1. Click the Trim-Winz Var button.
2. Select the By observation numbers option to mark an area according to
observation number.
3. Vertically mark the area of the plot to trim or Winsorize. The area becomes
yellow and the area outside the limits remains white. The Trim-Winz Var
dialog is updated according to the marking displaying the limits, percentages,
and the number of marked data points.

4. Click Apply.
Trim-Winz All dialog
Trim-Winz All is used with large datasets where it would be impractical to trim one
variable at a time.

View
417
To trim-Winsorize all variables at the same time:
1. In the Upper limit and Lower limit sections select the Metric to cut by: Std
Dev., Value, or Percent.
2. In Cut-off value field enter the value to trim or Winsorize by. Anything
outside these limits will be affected in the trim-Winsorizing.
3. In the Replacement options box, select the metric or how to replace the
affected values: Std.Dev., Value, Percent, Last Good Value, or Delete
Obs. When the selected replacement type is Std.Dev., Value, or Percent
enter values in the New value field. For more, see the table in the Defining
upper and lower limits By values subsection previously in this chapter.
4. Click OK.
Trim-Winsorizing all variables example
For example, to Winsorize the upper and lower 1% of the dataset and have the values
above and below the 1% replaced by the values at 1%, the dialog should look as
follows.

Click OK and the following message is displayed:

User Guide to SIMCA
418
Trim Winsorizing a selection of variables
To trim-Winsorize a selection of variables at the same time:
1. Make the selections in the Trim-Winz All dialog as described in the Trim-
Winz All dialog subsection previously in this chapter.
2. Click the Options button and select the desired subset of variables.
3. Click OK.

Trimming Winsorizing using limits and values from secondary
IDs
When using Trim-Winz All, limits and new values can be defined from variable
secondary IDs.
Note: When selecting cut-off values or new values from secondary IDs, the
corresponding fields in the main dialog become unavailable.
1. In the Trim-Winz All dialog, click the Options button.
2. Click the Limits tab.
3. There are four sections: Upper limit, New upper value, Lower limit, and
New lower value. Under each section select one of the options:
a. Defined in main dialog to define the limit or new value in the
Trim-Winz All dialog.
b. Secondary Variable ID and then select the secondary ID holding
the cut-off values or new values.
View
419

4. Click OK.
5. Select type metric of the cut-off-value and optional limits and new values
In the Trimming-Winsorizing dialog, the Cut-off value and New value
fields specified by a secondary variable ID are unavailable. The Metric and
Replacement options always have to be specified as well as values left with
Defined in main dialog selected.
For info on how to add secondary IDs to the dataset, see the Adding secondary IDs
section in Dataset spreadsheet section in the Home chapter.
Trim-Winsorizing using secondary ID example
In this example the dataset holds two secondary variable IDs named
PERCENTCutOffAndNewValueLOWER and VALUECutOffAndNewValueUPPER.
PERCENTCutOffAndNewValueLOWER contains the lower cut-off value and new
lower value in percent, 2%. VALUECutOffAndNewValueUPPER contains the upper
cut-off values and new values in actual values.

To use the secondary IDs as limits and new values:
1. Click the Options-button, and then the Limit tab.
2. Select the IDs as shown in the dialog and click OK.
User Guide to SIMCA
420

3. In the Trim-Winz All dialog:
- For the upper limit, select Value both as Metric and in the
Replacement options box. This specifies to the Quick Info to
interpret the Upper limit and New upper value from the
secondary ID as values.
- For the lower limit, select Percent both as Metric and in the
Replacement options box. This specifies to the Quick Info to
interpret the Lower limit and New lower value from the
secondary ID as percent.

4. Click OK and a summary message is displayed.

View
421
Variables window
The Variables window displays the variables included in the workset of the active
model.
Right-clicking the window opens the shortcut menu holding the following commands:
X, Y, Exclude, Select All, and Variable Label. These commands work as described in
the Marked Items subsection previously in this section.

Full Screen
To maximize the plot area, leaving only the minimized ribbon and the plot area, on the
View tab, click Full Screen. To return to normal, again click Full Screen.
Add to Favorites
Clicking Add to Favorites adds the active plot or list to the Favorites window. For
more, see the Favorites section earlier in this chapter.
Add to Report
Clicking Add to Report adds the active plot or list to the HTML Report. For more, see
the Generate Report section in the File chapter.
Window group
The Window group contains the regular window commands Cascade, Tile, and Close
Window. Clicking the dialog box launcher opens the Arrange Windows dialog.


423
Plot and list contextual tabs
The features available in the Plot and List, contextual tabs are context sensitive
appearing when applicable.
The Plot contextual tabs Tools and Layout become available when opening a plot and
display the features applicable to the active plot.
The List contextual tab Tools becomes available when opening a list and displays the
features applicable to the active list.
Marking in a plot or list opens the Marked Items contextual tab.
The contextual tabs include features like:
- The tools: Select, Zoom, Zoom Out, and Highlight Series.
- The properties: Color by, Size by, etc.
- The Change Type group features: Scatter, Column, List, Sources of
Variation, Out of Control etc.
- The layout features: loading and saving of templates, show/hide header,
footer, legend etc.
- After marking items: Exclude, Include, Labels, etc.
User Guide to SIMCA
424
Mini-toolbar - format plot
The mini-toolbar is available to customize the attributes of the plot. By default the
mini-toolbar is displayed when you click and then move the cursor northeast. When
right-clicking the mini-toolbar is displayed above the shortcut menu.
By clicking an item, line, symbol, text, etc., you can change attributes such as line
width, symbol type, color of a category, font in a text box, etc.


Note: When clicking a symbol or column, that symbol or column is marked
and the mini-toolbar for marked items appears.
Features available when clicking in plots are described in the table.
Click in plot area Mini-toolbar features left to right
An empty area in the plot
area.
Format Plot dialog, Properties dialog and zooming.
A symbol. Symbol style, size and color for the marked symbol
only. To change for all symbols in a series, use the
legend.
A line. Format Plot dialog, line width, style, and color of that
specific line.
A column. Color for the marked column only. To change for all
columns in a series, use the legend.
A limit. Format Plot dialog, line width, style, color and fill
color of that specific line. The last button allows
saving the current settings as default for the limit
type.
Note: After marking in a plot and making a change using the mini-toolbar, a
new series named Custom is created in the Styles node in the Format Plot
dialog, allowing further modification of the marked group.
Features available when clicking in the legend are described in the table.
Click in legend Mini-toolbar features left to right
Symbol
Format Plot dialog, hide series, symbol style, size
and color.
Line
Format Plot dialog, hide series, line width, style, and
color of that specific line.
Column Format Plot dialog, hide series, color.
Color after category coloring Format Plot dialog, color.
Plot and list contextual tabs
425
Click in legend Mini-toolbar features left to right
Color with continuous
coloring
Format Plot dialog, color 1 (start color), color 2 (end
color).
Features available when clicking in a text box are described in the table.
Click in text box Mini-toolbar features left to right
Any text box (header, footer,
axis title, etc.)
Format Plot dialog, hide selected, font face, font
size, grow font, shrink font, bold, italic, font color and
save as default style.
In the figure below both the legend mini-toolbar and the marking mini-toolbars are
displayed. To the right is the Color by feature floating after being torn off.

Managing the mini-toolbar
The mini-toolbar is always displayed above the shortcut menu, but you can specify
when you want to see it after clicking or marking in plots.
To hide the mini-toolbar that is displayed when marking:
- Click the x in the mini-toolbar or
- Click the cogwheel in the mini-toolbar that shows above the shortcut menu
and clear Show mini-toolbar on selection.
To hide the mini-toolbar that is displayed when clicking an item in a plot or in the
legend without marking, click the cogwheel in the mini-toolbar that shows above the
shortcut menu and clear Show mini-toolbar.
User Guide to SIMCA
426
Tools tab
The Tools contextual tab holds:
- The Change Type group with Scatter, Line, Column, List, and optionally
Out of Control Summary/Sources of Variation/Merge List.
- The Data group with Model, Comp, Batch etc depending on open plot/list
or window.
- The Properties group with Color by, Labels, and Size by.
- The Highlight group with the Select tool, Find, and highlight Series. Sort
is also positioned in this group.
- The zooming tools.

Change Type
In the Change Type group you can select to display the open plot or list as another
plot type or create a list. When changing between the regular plot types the plot
window is reused.
The possible buttons are:
- Scatter - Changes the current plot into a scatter plot.
- Line - Changes the current plot into a line plot.
- Column - Changes the current plot into a column plot.
- List - Creates a list of the content of the current plot. For Dendrogram see
the Create List for the HCA dendrogram subsection. See also the
Creating lists subsection in the Marked Items tab section.
- Out of Control Summary Plot (OOC) - described in the Out of Control
Summary Plot (OOC) subsection in this section.
- Batch Control Chart is available for the Out of Control Summary Plot
and displays the batches displayed in the OOC plot in a batch control chart.
- Sources of Variation Plot - described in the Sources of Variation Plot
subsection in this section.
- Merge List - available for the dendrogram plots.
- Other plots dialog dispatch launcher - opens the Create Plot dialog
described in the Creating plots subsection in the Marked Items tab section.
Note: With any plot or list open you can create another plot or list using the
same data by clicking Create | Plot on the shortcut menu.
Plot and list contextual tabs
427
Out of Control Summary Plot (OOC)
The Out of Control Summary Plot, OOC plot, displays a normalized column plot for
each batch that deviates from the limits in the current batch control chart, by
integrating the area outside the selected limits, for the selected phase and component.
The Y axis is in units of percent of the area outside the limits.
The Out of Control Summary (OOCSum) for all the vectors is always computed on
aligned vectors.
Creating the out of control summary plot
To open the OOC plot for the open batch control chart, on the Tools tab, click Out of
Control Summary Plot.

After creating the OOC plot you can recreate the batch control chart holding only the
batches that were out of control by, clicking Batch Control Chart on the Tools tab.

OOC plot example
From the batch control chart the Out of Control Summary Plot was created.

User Guide to SIMCA
428

In the Out of Control Summary Plot above, batch 28 has 20% of its area outside the
limits.
Creating Out of Control Summary Plot from Plot/List tab
The OOCSum vectors can be plotted and listed from the Plot/List tab by:
1. Clicking a plot type.
2. Selecting Batch Vectors in the Select data type box.
3. Adding the desired OOCSum vector.
See also the Batch Vectors - Out Of Control Summary subsection in Appendix A:
Statistics.
Contribution plot from OOC plot
Double-click any batch column in the Out of Control Summary Plot to display its
contribution plot.
Double-clicking a variable in the contribution plot opens the variable batch control
chart.

Plot and list contextual tabs
429
Sources of Variation plot
In batch level models it is useful to display the contribution and loading plots as line
plots over time rather than column plots at every time point. Therefore the Sources of
Variation plot is the default loading and contribution plot for all BLM with one or
more score or raw variables. The Sources of Variation plot contains the exact same
data as a loading or contribution plot would show. Instead of having the variable
number on the x-axis, the maturity of the batch is used. This gives a better view of how
the process variables relate to each other at different stages in the process.
For projects with phases, this plot is displayed showing all phases.
BLM with batch conditions
Variables without maturity, i.e. batch conditions, cannot be shown in Sources of
Variation plots. When the BLM contains batch conditions it can be useful to view the
plot as a regular column plot too.
To switch the Sources of Variation plot to a column plot, on the Tools tab, in the
Change Type group, click Column. A normal loading plot can always be created from
the Loading plot gallery, in the Diagnostics & Interpretation group on the Home tab,
by selecting one of the standard options.
Contribution plots in column form may be selected on the Marked Items tab, in the
Drill Down group, by clicking Column in the comparison plot gallery.
BLM loading and contribution plots example
How to change between a Sources of Variation and a column plot is described in the
table. This example uses a loading plot, but is applicable for contribution plots as well.
Step Illustration
1. On the Home tab, click
Loadings.

This plot displays how the selected variables
variation over time differs from the average of
batches or from another batch.
User Guide to SIMCA
430
Step Illustration
2. To transform to the
column plot, on the Tools
tab, in the Change Type
group, click Column.

3. To transform to the Sources of Variation plot, on the Tools tab, in the Change
Type group, click Sources of Variation.
4. To switch phases or displayed variables in the contribution plot open Properties
and:
- click the desired phase in the Select phase box.
- add to the Selected list the variables to display.
Create List for the HCA dendrogram
With the HCA dendrogram plot open, click List to display all clusters and the
calculated distances.


Plot and list contextual tabs
431
Create List for PLS-Tree
With the PLS-Tree dendrogram plot open, clicking List, creates a list displaying
details about the PLS-Tree sub-models.

Merge List for HCA dendrogram
With the HCA dendrogram open, Merge List is available in the Change Type group
on the Tools tab.

This command creates a list of the calculations behind the dendrogram.

User Guide to SIMCA
432
Data group
The Data group holds information about the active plot or list. These combo boxes
allow you to switch between models, components, variables etc.


Property
available
Description
Model The model of the current window. Lists all models in the project.
X-Axis
Comp, Y-
Axis Comp,
Z-Axis
Comp.
The component displayed on 1D, 2D, and 3D plots, such as Score
Line Plot, Score Scatter Plot, and Score Scatter 3D Plot.
Comp Component in the plot.
Available for cumulative plots and DMod plots.
X-Variable Lists all x-variables in the active model with the displayed x-
variables check boxes selected when more than one variable is
displayed.
Y-Variable Lists all y-variables in the active model with the displayed y-
variables check boxes selected when more than one y-variable is
displayed.
Observations Lists all observations in the active model, dataset, or predictionset
with the displayed observations check boxes selected when more
than one observation is displayed.
Batch Displays all batches in the active model or predictionset with the
currently displayed batches check boxes selected. When all check
boxes are selected, All Batches is displayed in the Batch box.
Some of the boxes in the Data group can be accessed using keystrokes. For more, see
the Switching components, batches, and models subsection after this section.
Plot and list contextual tabs
433
Switching components, batches, and models
Switching components on the plot axes, batches displayed, and model for the active
plot is available using keystrokes shortcutting to the Tools tab, Data group.
For regular projects, switch:
- Components on the Y-axis by pressing the UP and DOWN arrow keys.
- Components on the X-axis by pressing the LEFT and RIGHT arrow keys.
- Models by pressing the PAGE UP and PAGE DOWN keys.
For batch projects, in the batch control charts switch:
- Components on the Y-axis by pressing the UP and DOWN arrow keys.
- Batches by pressing the LEFT and RIGHT arrow keys.
- Phases (models) by pressing the PAGE UP and PAGE DOWN keys.
The active window and property bar are updated according to the changes.
Properties group
The Properties group, on the Tools tab, contains the commonly used Properties
dialog features: Color by, Labels, and Size by.
Note: Only a selection of coloring types is available from Color by. The Color
page in Properties contains more coloring options.
The Properties dialog is opened by clicking the dialog box launcher in the Properties
group on the Tools tab.

Color by on the Tools tab
Color by is available for plots displaying observations or variables.
Available are for example:
- Default coloring - all points are colored the same color.
- Color by Observation ID/Variable ID - the available secondary IDs are
available.
- Color by Variable - for observation plots the variables of the included
datasets are listed.
- Color by Batches - for batch control charts.
- Color by Classes - for observation scatter plots where the model has more
than one class specified.
Note: Color by can be torn off and docked by pulling the dotted part.
User Guide to SIMCA
434

Color by torn off from the ribbon and floating.

Label Types for plots
On the Tools tab, you can select the label type for the symbols in the current plot by
clicking Labels in the Properties group.

In the Properties dialog for plots, in the Label Types page, the labels to use as Point
Label, Axis Label, and Title label can be specified.
To specify a Point Label:
1. Select the primary or secondary identifiers in the Use label box.
2. Enter the character in the label to start with in the Start field.
3. Enter the number of characters to use in the Length field.
To specify an Axis Label:
1. Select the primary or secondary identifiers in the Use label box.
2. Enter the character in the label to start with in the Start field.
3. Enter the number of characters to use in the Length field.
Plot and list contextual tabs
435
To specify the Title labels:
1. Click the Title labels box.
2. Select the desired variable ID.
Note: The variable ID selected in the Title labels box is displayed in all
applicable headers, footers, legends, and axis labels.

Label Types for lists
On the Tools tab, in the Properties group, you can select to Display all labels or a
specific label in the list using Labels.

In the Properties dialog for lists, in the Label Types page, the following options are
available:
1. Displaying a specific ID by selecting Use identifier and
a. Selecting an identifier in the list. By default the primary ID.
b. Enter the character in start with in the Start field. By default 1.
c. Enter the number of characters to use in the Length field. By
default 10.
2. Selecting Display all identifiers resulting in displaying all available
identifiers in their full length.
User Guide to SIMCA
436

Size by
The plot marks of all scatter plots, both 2D and 3D, can be sized according to a
selected vector. Additionally, all scatter plots of vectors of the type variables and
scores can be enlarged when the observation is found outside the D-Crit of DModX or
DModXPS.
Sizing is available on the Tools tab, in the Properties group, by clicking Size by.

Plot and list contextual tabs
437
Sizing by a vector
When selecting to Size by Vector, the Properties dialog opens and all applicable
vectors are available.
For instance, for a score plot, all vectors found with data type Variables and Scores in
the Plot/List | Scatter dialog are available.
When clicking OK, the size of the plot marks will be smaller or larger than the default
plot mark depending on the value for the specific observation or variable in the
selected vector.

Enlarging when outside D-Crit
When clicking Size by | Enlarge when outside D-Crit, all observations with DModX
outside D-Crit are enlarged.

Enlarge when outside D-Crit is only available for plots displaying observations.
To shrink or enlarge all plot symbols, use Symbol Style page in Format Plot.
Properties dialog
All plots and lists have a context sensitive Properties dialog. Many of the pages in this
dialog are general and available for several plot types. These pages are described in
this section. Other pages are plot specific and are therefore described in connection
with the description of the plot.
The table describes the different ways to open the Properties dialog.
User Guide to SIMCA
438
Open Properties by Screenshot
Clicking the plot (in an area where there
is no points) and clicking Properties in
the mini-toolbar

Clicking the dialog box launcher in the
Properties group on the Tools tab

Right-clicking the plot and clicking
Properties from the shortcut menu.

The property pages described in this section are:
- Color
- Label Types for plots - earlier in this section.
- Label Types for lists - earlier in this section.
- Size - earlier in this section.
- Item Selection
- Limits
- Components
- Select Y Variable
- Number Format
Color in the Properties dialog
With all plots displaying variables or observations, the plot symbol, line, or column
can be colored as described in the table, by clicking Color by on the Tools tab.
Clicking More options opens the Properties dialog with the Color tab open. Here all
coloring types for the active plot are available.
Note: When coloring a line plot, the connecting line when the line changes
color starts with one color and gradually turns into the other. However, when
a point in a line plot does not have a connecting line on either side, the plot
mark of that point is displayed.
Plot and list contextual tabs
439
This table describes the coloring types available in the Color page and how to specify
after selecting the respective coloring in the Coloring type box.
Note: The color of the classes/batches/categories etc can be changed in
Format Plot, Styles, Category Coloring.
Color by Action
Batches Plots created for batch evolution models can be colored by Batches.
When coloring by Batches, each batch is then assigned a color that it
keeps independent of which batches are displayed in the plot.
Classes When classes are present in the plot, the observations are by default
colored by Classes.
To change the colors of the individual classes, use the legend mini-
toolbar, or Format Plot, Styles, Category Coloring.
Density When a plot has many observations, coloring by Density (density
function) is sometime useful.
When coloring by Density, the plot marks are colored according to
how many points are positioned close to them. This means that, using
the automatic color scheme, a point with many points close is colored
red while a point with few points close is colored blue.
Available for plots displaying variables or observations.
To change the start and end colors, use the legend mini-toolbar or
Format Plot, Styles, Coloring.
Identifiers To color by Identifier:
1. Select the identifier in the Choose the ID to color by box. By
default the first secondary ID is selected.
2. In the Start and Length fields enter values to specify which
characters in the ID to use.
Available for plots displaying variables or observations.
To change the color, use the legend mini-toolbar or Format Plot,
Styles, Category Coloring.
Observation
Classes
When classes have been specified, each line of the XObs, YObs and
Spectra plots can be colored by the class belonging of the observation.
Observation
Identifiers
Each line of the XObs, YObs and Spectra plots can be colored by the
observation ID. For how to specify, see the Identifiers description
above.
Observation
Vectors
Each line of the XObs, YObs and Spectra plots can be colored by an
observation vector. For how to specify, see the Vector (continuous)
description below.
Marked
Groups
When observations are marked, you can color by Marked Groups.
When coloring by Marked Groups the observations in the first group
are displayed in one color, the observations in the second group in
another color, etc.
Available for plots displaying variables or observations and especially
useful after marking using the Group Dendrogram Clusters tool.
User Guide to SIMCA
440
Color by Action
Models and
Predictions
Prediction plots can be colored by Models and Predictions. This
means that observations belonging to different models are colored in
different colors and the observations that are not part of a model are
colored in yet another color.
Useful in the Coomans' plot.
Patterns To color by parts of identifiers, by Patterns:
1. Select the identifier in the Choose the ID to color by box. By
default the first secondary ID is selected.
2. Select Begins with, Contains, Ends with, or Equal to in the
Pattern box.
3. Enter the characters to color by.
4. Click the Add Pattern button.
5. Repeat 1-4 until done.
Available for plots displaying variables or observations.
Predictions Observations that are part of the workset are colored in one color, and
observations that are only part of the predictionset are colored in
another color.
Terms Plots displaying variables can be colored by Terms. This results in that
all types of terms (original X, expanded terms square, cross, cubic,
etc.) are colored in different colors.
Color by Terms is the default for the loading scatter plot.
Variable
Classes
When variables have been assigned to blocks, use Variable Classes
to color these blocks.
Vector
(Categories)
To color by Categories:
1. Select data source and variable in the Data respective Item
boxes. Note here that both model terms and variables are
available.
2. Optionally enter the desired number of groups in the Split range
field.
3. Click the Add Category button.
4. The range of the added categories can be modified by clicking
the category and modifying the fields beneath the Remove All
button.
Note: To color by batch conditions or hierarchical variables, select that
dataset in the Data box. To color by model results, select the model in
Data and the vector in Item.
Vector
(Continuous)
To color by Vector (Continuous), select data source and variable in
the Data respective Item boxes. Note here that both model terms and
variables are available in Item.
Note: To color by batch conditions or hierarchical variables, select that
dataset in the Data box. To color by model results, select the model in
Data and the vector in Item.
Plot and list contextual tabs
441

SIMCA by default remembers the selected coloring when opening the next plot.
For more, see the More Options subsection in the SIMCA Options section in the
File chapter.
Item Selection
To remove some items from the current plot or list:
1. Mark the points in the plot.
2. Click Hide on the Marked Items tab.
As a result the Item Selection page in the Properties dialog is updated accordingly.
The Item Selection page displays the Selected and Unselected items. Removing and
adding items from the current plot or list can also be done here.
1. Select the items in the Selected list.
2. Click <= to move them to the Unselected list.
To display a different identifier, select another ID by clicking the ID box.

For more about the Find feature, see the Find feature in workset dialog subsection in
the Workset section in the Home chapter.
Limits
The Limits page enables customizing the displayed confidence intervals, limits, or
ellipse.
User Guide to SIMCA
442
The following plots have a Limits page in their Properties dialog for the described
limits:
Plot Limit Option
Score
Scatter Plot,
2D and 3D
Hotelling's
T2 ellipse
T2 Ellipse box with: Display (default) and Hide.
Significance level field with default value 0.05.

Score Line
Plot
Standard
deviation
lines
SD Limits box with: Display (default) and Hide.
Customize SD Limits field with the default -3 -2 0 2
3.
Note that all limits inside the range -3 to 3 are
colored in green, and all outside red.
'0' represents the average.

Score and
Loading
Column
plots,
Coefficient
Plot, and
VIP.
Jack-
knifing
uncertainty
bars on the
columns
Confidence level box with: Default (the setting in
Model Options), None (to not display the error
bars), 99%, 95%, and 90%.

Hotelling's
T2Range
95% and
99% T2Crit
T2Crit Limits box with: Display (default) and Hide.

Distance to
model
D-Crit D-Crit Limit box with: Display (default) and Hide.
Significance level field with default value 0.05.

Plot and list contextual tabs
443
Plot Limit Option
Y Predicted
Column Plot
Jack-
knifing
uncertainty
bars on the
columns
YPredPS Confidence Interval box with: Display
(default) and Hide.

Coomans'
Plot
D-Crit D-Crit Limit boxes with: Display (default) and Hide.


Components in Properties
For a number of plots, the Properties dialog has a Components page.
Use this page to select for which components to display the plot.
For instance, selecting another component than Last Component for the Coefficient
Plot will display the plot cumulative including the selected component.
For the Hotelling's T2Range Plot, both the From component and To component has
to be selected.
Note: When adding a plot to Favorites or the report, Last Component is often
preferable.
Select Y Variable
The Coefficient Plot, Residual Normal Probability Plot, Observed vs. Predicted
Plot, and all Y Predicted plots found in the Y PS gallery on the Predict tab, are all
displayed for one y-variable at a time. By default the plots are displayed for the first y-
variable.
User Guide to SIMCA
444
To switch to another y-variable in the plot use one of the following methods:
- Click the Y-Variable box in the Data group on the Tools tab, and select
the desired y-variable.
- Right-click the plot, click Properties, click the Select Y Variable tab,
and select the new variable in the Select Y Variable box.
Number Format
Selecting the number format type and precision applies only to lists.

The available format types in the Type box are:
- Default - displaying the numbers in decimal format with the number of
digits stated in the Precision box but leaving ending zeros hidden.
- Exponential - displaying the exponential format using the stated Precision.
- Decimal - displaying the number of decimals stated in the Precision box.
The values can be displayed using 1-6 digits. This is selectable in the Precision box.
Select - Marking tool
The Select-button is used to select which type of marking to use.
Click the small arrow below the button and click the type of marker from the menu.
The available marking types are:
- Free-form selection: Allows marking to take any shape.
- Rectangular selection: Allows marking in a rectangular shape.
- Select along the X-axis: Marks as a vertical bar.
- Select along the Y-axis: Marks as a horizontal bar.
- Batch Marking Mode: Marks all observations belonging to the selected
batch. Using the Exclude button after marking an entire batch excludes the
selected batch from all the models (phases). A new BM model is
automatically generated with a BEM holding a PLS model for each phase.
- Group Dendrogram Clusters: Available for an open Dendrogram and
displays a horizontal line with which the resolution of the dendrogram is
specified, and thereby the number of clusters. For more, see the Group
Dendrogram Clusters marking tool subsection in the Analyze chapter.
Plot and list contextual tabs
445
Note: The Batch Marking Mode is an on/off switch to marking entire batches
instead of points. The marking type is still Free-form selection, Rectangular
selection etc.
The table displays the available commands from the Select plot items tool.
Menu Available for

Regular project standard plots.

Batch evolution models standard plots.

Dendrogram plots open in any project.
Note: To keep the marked points marked, while marking new points, hold
down CTRL. The marked points then belong to the same group of marked
items.
Deselecting/unmarking points
To deselect points in a plot, click a white area in the plot.
To deselect points in a list, mark the first row.
See also the Marked Items tab section later in this chapter.
Displaying properties of the item
When positioning the cursor on an item in a plot, the marker behaves as a pointer and
displays the name and coordinates of the observation or variable.
User Guide to SIMCA
446
Creating contribution plots
With any marker:
1. Double-click an observation in a score plot for instance. The contribution
plot opens, comparing the variables of the selected observation to the
average of all the observations in the workset. This plot indicates why the
selected observation is different from the average.
2. Click the first observation, and then double click the second observation. The
resulting contribution indicates why the first observation differs from the
second.
For more about creating contribution plots, see the Creating plots from plots section
later in this chapter.
Selecting points according to color in legend
All points in a color can be selected by clicking that color in the legend.

Find
With an active plot, list or spreadsheet, Find enables finding items meeting selected
criteria.
The Find dialog is opened by clicking Find in the Highlight group on the Tools tab,
by pressing CTRL+F or from the Quick Access Toolbar.

Plot and list contextual tabs
447

In the table, the available items in the Find in, Find what, and Logical Expression
boxes are listed.
Find in Searches in Logical expressions
Identifiers The text or values of the selected
identifier, primary or secondary, or all
identifiers.
For batches the secondary identifiers
include the batch level IDs Phase,
Maturity, Source, and Number, and
the batch project IDs $BatchID and
$PhaseID when available.
For numerical identifiers:
>, >=, <, <=, between, =,
not equal.
For text identifiers:
begins with, ends with,
contains, =, not equal.
All Data The vectors displayed on all axes. For
the default score scatter plot, the
search is done on both t1 and t2.
>, >=, <, <=, between, =,
not equal.
Series
Data
The vectors that are selected as
series. For 2D scatter, line, and
column plots that means that the
search is limited to the y-axis vectors
while for the 3D scatter plot the search
is limited to the z-axis vectors. For
example, for the default score scatter
plot, the search is limited to searching
in the vector t2 found on the y-axis.
For lists, this option is the same as All
Data.
>, >=, <, <=, between, =,
not equal.
X Axis The vector displayed on the x-axis.
Available for plots only.
>, >=, <, <=, between, =,
not equal.
Vectors The selected vector in Find what.
When selecting Vectors in Find in:
- The Find what box contains all
vectors available for the type of
vector displayed in the active
plot or list, e.g. if the plot
displays t1 vs. t2, the available
vectors are all vectors of data
type Variables and Scores.
- The Vector Details section
becomes available enabling
selection of model.
>, >=, <, <=, between, =,
not equal.
User Guide to SIMCA
448
Find in Searches in Logical expressions
Text The list as if all entries where text and
marks the matching cells. Available for
lists and spreadsheets only.
begins with, ends with,
contains, =, not equal.
Note: To search in a variable, in the Find in box select Vectors and in the
Find what box select XVar. Then in the X Terms box in the Vector details
section, select the desired variable.
Highlight Series
With highlight Series clicked, hovering over one series in the legend grays all series
but the hovered one.

Sorting ascending or descending
All lists, the dataset spreadsheets, and column plots can be sorted ascending or
descending.

Sorting lists
Sorting of the dataset spreadsheet is available both as sorting the presentation and as
sorting the dataset. Sorting is only available when one variable is marked.
To sort:
1. Mark the variable to sort by.
2. On the Tools tab click Sort | Sort ascending or Sort descending.
3. When sorting a dataset, in the Sort Dataset dialog that opens select:
a. Sort the list to sort only the presentation of the data in this
spreadsheet. This sorting will not delete the models, only sort the
list. When creating a new workset, the original order of the
observations will be used.
b. Sort the dataset to sort the dataset the models are built from. With
this option all models will be deleted. When creating a new
workset, the new sorted order of the observations will be used.
Note: Read-only datasets cannot be sorted.
Plot and list contextual tabs
449
Sorting column plot by value or ID
All column plots, including contributions, can be sorted by strings in the primary or
secondary ID and/or by values. This allows displaying variables grouped as desired.
The various groups will have different colors.
To sort a column plot:
- On the Tools tab, in the Highlight group, click Sort.
- Alternatively, right-click the column plot and then click Sort Ascending or
Sort Descending.
To sort a column plot by ID
1. On the Tools tab, in the Highlight group, click Sort | Advanced sorting.
2. Select how to sort the data in the Sort Items dialog and then click OK.

The Sort Items dialog is described in the table below.
Option Description Default
Sort by ID Select the Sort by ID check box to sort
the column plot by the selected ID.
After selecting the Sort
by ID check box, the
primary ID is the default.
Start To sort by a part of the ID, enter the
start character position in the Start
field.
'1' which means that the
starting point is the first
character.
Length To sort by a part of the ID, enter the
number of characters to use in the
Length field.
By default the entire ID
is used in the sorting,
that is Length is
<empty>.
Color by
identifiers
Select the Color by identifiers check
box to color the columns according the
identifier selected to sort by.
Not selected.
Sort by Selecting the Sort by check box and in
the Sort by box select value or
absolute value to sort the column plot
accordingly. Selecting to sort by
absolute value results in sorting
according to size of the column,
independent of whether it is positive or
negative.
After selecting the Sort
by check box, value is
by default selected.
User Guide to SIMCA
450
Option Description Default
In series When selecting the Sort by check box,
the sorting has to be done using one of
the series in the plot. When available
you can select another series if
desired.
After selecting the Sort
by check box, the first
series is by default
selected.
Zooming in
To zoom in a plot, click the Zoom tool in the Zoom group. By default Zoom In is
selected.

Use the Zoom menu to select the type of zoom from the drop down menu:
- Zoom In: Magnifies a rectangular region
- Zoom X: Expands the x direction
- Zoom Y: Expands the y direction
- Zoom subplot: Magnifies a subplot in a multi-plot display.
Then mark the desired region, or subplot, of the plot to zoom.
Note: With any zooming type selected, double-clicking a subplot
automatically zooms the subplot.
Zooming in a scatter 3D plot is described in the Zooming in 3D scatter plot
subsection in the Score Scatter 3D Plot section in the Home chapter.
Zooming out
Click Zoom Out to revert zoom to original scale in the steps taken when zooming.
Zoom Out is available for all lists and spreadsheets from the shortcut menu.
Layout tab
The Layout contextual tab holds features pertaining to how the plot is displayed.

The Templates group holds the features concerning the formatting in the plot.
Plot and list contextual tabs
451
The Show group controls which texts outside the plot are that are displayed, plus the
Regression Line (for 1D and 2D plots) and the Resolve Coefficients for hierarchical
top model coefficient plots.
Format Plot to the right on the tab opens the Format Plot dialog.
Templates group
The Templates group holds features pertaining to the formatting of plots. This tab is
available when the active window is a plot.
The following features are available:
- Load described in Switching plot formatting templates subsection.
- Save as and Save as Default described in Saving Format Plot template
subsection.
- Open templates folder - opens the folder holding the templates when
clicked.
- Restore Default described in Restoring to default plot
formatting subsection.
The Format Plot is positioned to the far right on this tab, but is described in the
Format Plot section later in this section.

Switching plot formatting templates
After having added customized plot formatting templates, these templates can be
selected by clicking the Load button.
After loading a new template, that template is applied to all open and future plots.
The template Default (optimized for speed) should be used when your project has so
much data that the plotting becomes slow. In this template the most CPU consuming
features have been turned off.

Saving Format Plot template
After customizing the plots using Format Plot you can save the settings of some
attributes, such as fonts, gridlines, symbol type etc, to a template.
To save a template, with the plot open, click Save as Default or Save as and specify a
name.

User Guide to SIMCA
452
Note: Only the formatting available in the current plot is saved to the
customized plot formatting configuration. All other formatting will remain
default. This means that formatting of headers, footers etc. apply to all plot
types (e.g. scatter, column, line) while default line color, default symbol shape
are plot specific and need to be specified with that plot type open.
Restoring to default plot formatting
To restore to the Umetrics default plot formatting template, click Restore Default.
Clicking Restore restores the Default plot formatting to the Umetrics default plot
formatting and switches to it.
Format Plot
The most common attributes of the plot axes, plot area and of the header and footer can
be customized directly in the plot using the mini-toolbar, or by opening the Format
Plot dialog.
The mini-toolbar is displayed after clicking an item (header, footer, legend, symbol,
column etc.) and moving the cursor northeast.
To open the Format Plot dialog use one of the following methods:
- On the Layout tab, click Format Plot (button to the far right).
- Double-click the plot area, header or footer, or axes in the plot.
- Right-click the plot and then click Format Plot.
Axis
The properties under the Axis node apply to the selected axis: Axis X, Axis Y, Axis
Y2 (second Y-axis) etc.

Plot and list contextual tabs
453
In the Format Plot dialog, Axis node, click:
- The respective axis to customize the scaling of the selected axis, annotation
rotation etc.
- Axis General to access the following pages:
1. Axis General to customize color, width and other properties of all
axes.
2. Tick Marks to select how to display major and minor tick marks.
3. Axis Font to customize the font of the axis annotation.
4. Title Font to customize the font of all axis titles.
Changing the scale, tick mark label and the axis properties
Use the General page of the Format Axes dialog to change the scale, tick mark label
and axis properties of the currently marked axis.

Individual axes
The content for the individual axes is described in the table below:
Field/button
name
Displays Result after entering a new
value and clicking Apply
Axis
Show axis Default selected. If cleared, that axis is not displayed.
Minimum Start point for the
selected axis when
the values are
displayed in regular
order. For time-
variables the page is
adjusted, see the
Time axis topic.
The new start point is used for the
selected axis.
User Guide to SIMCA
454
Field/button
name
Displays Result after entering a new
value and clicking Apply
Maximum End point for the
selected axis when
the values are
displayed in regular
order.
The new end point is used for the
selected axis.
Auto adjust
scales for
suitable limits
and step size
Default selected. If cleared, enables the Step size
field.
Step size Default disabled. The entered value defines the major
tick mark spacing for the selected
axis.
Reverse axis Default cleared. If selected, the scale is displayed
with the highest number to the left
and the lowest to the right.
Annotation
Rotation Default No Rotation. Selecting 90 degrees displays the
annotation turned 90 degrees.
Title
Show title Default selected. If cleared, the axis title is not
displayed.
Title The vector name. The new title. Title changes cannot
be saved.
Note that the distance between the tick marks is constant.

When you have more than one series in a plot, you can select to display more than one
y-axis. For more, see the Multiple Y axes subsection later in this chapter.
Plot and list contextual tabs
455
Time axis
The content for the Axis X and Axis X Title pages for a time variable is described in
the table below:
Field/button
name
Displays Result after entering a new
value and clicking Apply
Show axis Default selected. If cleared, that axis is not displayed.
Minimum Start point in time
units. Change by
typing in the field.
The new start point is used for the
selected axis.
Maximum End point in time
units. Change by
typing in the field.
The new end point is used for the
selected axis.
Auto adjust
scales for
suitable limits
and step size
Default selected. If cleared, enables the Step size
field.
Step size Default disabled. The entered value defines the major
tick mark spacing for the selected
axis.
unit The unit of the step
size.
Switching between units updates the
step size automatically.
Axis X Title tab
Rotation under
Annotation
Default No Rotation. Selecting 90 degrees displays the
annotation turned 90 degrees.
Show title
under Title
Default selected. If cleared, the axis title is not
displayed.
Title The vector name. The new title. Title changes cannot
be saved.
Note that the distance between the tick marks is constant.

User Guide to SIMCA
456
Axis General
In Axis General the following features are available and are applied for all axes:
- Axis Color and Width.
- Autoscale - when selected it autoscales all axis. Clearing it makes the current
ranges of the axes stick even if the vector displayed is switched.
- Show arrows - when selected the arrows are displayed at the end of the axes.
- Arrow Color - Automatic here means the same as the axis color.
- Annotation Color and Distance from axis.
- Axis title Color.

Tick Marks
Use the Tick Marks page to specify how the tick marks should be oriented and how
long they should be.
Under Major tick marks and Minor tick marks the same boxes are available:
- Tick mark type - where None is no tick mark, Outside is to have the tick
mark outside the axis, Inside is to have the tick mark inside the axis and
Cross is to have the tick mark on both sides of the axis.
- Tick mark size is the length of the tick mark.

Plot and list contextual tabs
457
Font
Use the respective font pages, to select the Font, Font style, Size, and Effects for the
annotation, title, etc.


Section Description
Font Displays the currently selected font. Click another font to switch.
Font
style
Displays the current font style. Select to display the text Regular
(default), Bold, Italic, or Bold Italic.
Size Displays the current size. Select another size as desired.
Preview
Displays a preview of the expected text according to the selections
above.
Anti
aliased,
Filled
check
boxes
Allows selecting to display the text Anti aliased and/or Filled or
neither.

User Guide to SIMCA
458
Customizing the gridlines
Use the Gridlines page to customize the gridlines and grid stripes. The gridlines are
placed on the major tick marks. Grid stripes are the areas between the gridlines, where
every second such area can be colored.
Note: Gridlines and grid stripes can be specified individually for Vertical and
Horizontal. Specification using the node applies to both.
The following is available:
Option/Box/Field Description
No gridlines No gridlines are displayed when No gridlines is
selected.
Gridline When selecting Gridline you can select to display the
lines Solid, Long Dash, Dotted, or Dashdot.
Gridline Color The color of the grid, by default gray.
Width The width of the gridlines.
No grid stripes No grid stripes are displayed when No grid stripes is
selected.
Grid stripe Solid fill When selecting Solid fill the grid stripes are displayed in
the selected color.
Grid stripe Color Displays the current color and allows selecting a new
color.
Grid stripe Gradient
fill
When selecting Gradient fill you can select to display
the stripes Horizontal, Vertical, Forward Diagonal,
Backward Diagonal, Radial, Horizontal Bar, Vertical Bar.
Grid stripe Color 2 Displays the current color and allows selecting a new
second color for Gradient fill.

Plot and list contextual tabs
459
Background
The Background node controls the attributes of the background displayed in the plot.
Note: The background fill and border can be specified individually for the
Window Area and Plot Area. Specification using the node applies to both.
The table describes the available options:
Option/Box/Field Description
Fill
No fill No background is displayed when No fill is selected.
Sold fill When selecting Solid fill the background is displayed in
the selected color.
Gradient Fill When selecting Gradient fill you can select to display the
background Horizontal, Vertical, Forward Diagonal,
Backward Diagonal, Radial, Horizontal Bar, Vertical Bar.
Color Displays the current color and allows selecting a new color
for the background.
Color 2 Displays the current color and allows selecting a new
second color for Gradient fill.
Border
No border No border around the background area is displayed when
No border is selected.
Solid line When selecting Solid fill, the background is displayed in
the selected color.
Color Displays the current color and allows selecting a new color
for the border.
Width The width of the border.

User Guide to SIMCA
460
Titles in Format Plot
Use the Format Plot dialog to customize the items available in the Titles node. When
customizing, the changes apply to the selected item,
Header/Footer/Timestamp/Subheader.
Note: Changing color, font and size of the
Header/Footer/Timestamp/Subheader can be done directly in the plot by
clicking and using the mini-toolbar.

On the Titles page:
- Select to hide or show by selecting or clearing the Is Visible check box.
- Select where to display the title in the Anchor box, when applicable.
- Customize the text displayed by changing or typing in the field.
- Align the text right, center, or right by clicking the appropriate alignment
button.
- Customize the color of the text in the Text color box.
- Customize the background color by clicking the Background color box.
- Select to display a border by selecting the Is visible check box in the Border
section. With it selected you can customize Color, Margin (to the text) and
Width of the border.
On the Font page you can select the Font, Font style, Size, and Effects. See also the
Font subsection earlier in this chapter.
Legend
The Legend page controls the attributes of the legend, such as placement, color, and
border.
Plot and list contextual tabs
461
The table describes the available options:
Option/Box/Field Description
Placement section
Show legend The legend is displayed when the Show legend check
box is selected.
Position By default the legend is positioned Top right. To change
from the default, click one of the available placement
options.
Orientation The orientation is by default Vertical. Horizontal is the
other option.
Text alignment Alignment of the text in the legend is by default Left.
Right and Center are the other options.
Color section
Text color By default the text is black.
To customize text color of the legend, click the Text
color box.
Background color The default background color is white.
To select another color, click the Background color
box.
Border section
Is visible By default the border is not displayed. To display the
border around the legend, select the Is visible check
box.
Color To customize the border color of the legend, click the
Color box in the Border section.
Margin To specify the margin to the text, change the value in the
Margin field.
Width To increase or decrease the width, enter a new number
in the Width field.

User Guide to SIMCA
462
Limits and Regions
There are a number of limits available in plots in SIMCA, for instance the ellipse in the
score scatter plot, the DCrit limit in the DModX plot etc. Additionally there are
regions, such as the background of phases in batch control charts and models in
hierarchical top model coefficient and contribution plots that are colored to
differentiate between models.
The attributes of the limits and regions can be modified individually for each limit and
region type under the Limits and Regions node in the Format Plot dialog, in the
respective Region Style pages. The Region Style page is described in this section.
For multi plots, Control Charts and Wavelet Structure, the limit pages are
positioned last in the Format Plot dialog under Plot 1, Plot 2, Plot 3, etc.
For batch control charts, the limit pages are positioned under Styles.
Customizing Limits and Regions
The selected limit or region can be customized in the Region Style page.
Feature Description
Line
Style Select between No line, Solid, Long Dash, Dotted, Dash Dot.
Width Increase or decrease as desired.
Color Displays the current color and allows selecting a new color.
Fill
Type Select to fill the area Over or Under the limit, or select No fill.
Style
Select the fill to be Solid or Gradient. Selecting Solid enables the Color
1 box and selecting Gradient enables the Gradient, Color 1 and Color
2 boxes.
Gradient
When selecting Gradient you can select to display the fill Horizontal,
Vertical, Forward Diagonal, Backward Diagonal, Radial, Horizontal Bar,
Vertical Bar.
Color 1 Displays the current color and allows selecting a new color.
Color 2
Displays the current color and allows selecting a new color. Available
when the Style is Gradient.
Font tab - see the Font topic.
Label Style - see the Labels topic.
Plot and list contextual tabs
463

To not display the limit, see the Limits subsection previously in this chapter in the
Tools section.
Label Style
The Labels pages control the attributes of the displayed plot marks. Color, alignment,
font and rotation apply to all labels in a given series.
The tabs available in the Labels node, Label Style, Font and General, are described
in the table.
Option/Box/Field Description
Label section in Label Styles
Position The default position of the label in reference to the point is
Right. To position the label in another direction, click the
Position box and make a new selection.
Offset The offset defines the distance between the label and what
it labels in the direction of the selected Position.
To increase or decrease the distance between the label
and item, enter a new value in the Offset field.
Rotation The default rotation of the labels is 0. To rotate the label,
enter a new value in the Rotation field. Note that this field
is unavailable when Avoid overlapping labels has been
selected in the General tab.
Color section in Label Styles
Text color Displays the current color of the text in the label and allows
selecting a new color. Automatic displays the text in the
same color as the symbol fill
Background color The background color of the label is default transparent. To
select a color, click the Color box and make the selection.
Border section in Label Styles
Is visible By default, no border for the label is displayed. To display a
border around the label select the Is visible check box.
User Guide to SIMCA
464
Option/Box/Field Description
Color Displays the current color of the border and allows
selecting a new color.
Margin The margin between the text and the border can be
customized in the Margin field.
Width To increase or decrease the width, enter a new number in
the Width field.
Draw connection
line
Draw connection line is by default cleared. Selecting it
will display a thin line between the label and its point.
Font tab - see the Font topic.
General tab
Avoid overlapping
labels
With Avoid overlapping labels selected, point labels will
try not to overlap each other. A high specified limit in the
Limit field in combination with many labels can be quite
time consuming.

Error Bars
The Error Bars page controls the attributes of the error bars displayed in the score,
loading, coefficient, VIP, and Y PS column plots.
The settings available are:
- Is visible - hides the error bars when cleared.
- Color of the error bars - by default black. To display the error bars in any
other color, click the Color box.
- Line width of the vertical and horizontal lines. To display wider error bars,
enter a new number in the field.
- Error bar width is how wide the horizontal line is compared to the column.
By default 60%.
Plot and list contextual tabs
465

Styles
In the Styles node the attributes of the series, Category Coloring and for BCC
plots also limits can be customized.

The available pages in the Styles node, different depending on which plot is open, are:
- Marking Style
- Category coloring
- Symbol Style
- Line Style
- Column Style
- Fill Style - See Region Style Fill section.
- Options
- Y-Axis
- Font
User Guide to SIMCA
466
Marking style
In the Marking Style page you can change the marking colors by marking and
selecting a new color in the color box to the right.

Category coloring
Category coloring is available after coloring and the Use different symbols on
category colored values check box in the Options page is not selected.
In this page, the colors can be changed as desired.

Plot and list contextual tabs
467
Symbol Style
The Symbol Style page controls the attributes of the symbols displayed in the plot.
The table describes the available options:
Option/Box/Field Description
Shape The shapes of the series can be changed by:
1. Marking a series.
2. Selecting a new shape in the Shape box.
With None selected, no symbols are displayed.
Size To increase or decrease the size, enter a new number in
the Size field.
After using the Size by feature, changing the value in Size
affects the points proportionally.
Fill
No fill Results in symbols transparent inside the outline.
Solid fill Displays the color selected in Color.
Gradient fill Allows you to select to display the symbol with a gradient
fill of type Horizontal, Vertical, Forward Diagonal,
Backward Diagonal, Radial, Horizontal Bar, Vertical Bar.
Color Displays the current color. To display another color, click
the Color box, and then select a new color.
Color 2 Displays the current second color for Gradient fill and
allows selecting a color.
Outline
No outline No outlining contour of the symbols.
Solid line The outline of the symbol is displayed.
Color Color of the outline. Automatic results in an outline in the
same color as the symbol but in a darker shade.
Width Width of the outline.
Glow
Use glow When selected an area outside the outline is colored for a
glowing effect.
Color Color of the glow.
Width Width of the glow.
User Guide to SIMCA
468

Line Style
The Line Style page controls the attributes of the lines displayed in the plot.
The table describes the available options:
Option/Box/Field Description
Pattern Change the pattern of the line by clicking the Pattern
box and selecting another pattern.
The available types are: Solid, Long Dash, Dotted, and
Dash Dot.
Width To increase or decrease the width, enter a new number
in the Width field.
Color To display another color, click the Color box, and then
select a new color.
Smoothed line
(Bezier)
Select Smoothed line (Bezier) to smooth out the edges
of the line.
Note: Filling is available in the Fill Style tab. See the Customizing limits
subsection for details.

Plot and list contextual tabs
469
Column and Column Style
The Column and Column Style pages control the attributes of the columns displayed
in the plot.
The Column page, found above the Styles node, displays the current Column width
and allows adjusting it.
The Fill and Border options are described in the Symbol Style subsection; Border is
there named Outline

Options page in Format Plot
Selecting the Use different symbols on category colored values check box will
display different symbols for the different colors when coloring a plot.
This also has the effect that each color becomes a series in the Format Plot dialog and
thus enables the Symbol Style features.
Note: Symbols for continuous coloring are not affected by this option.

User Guide to SIMCA
470
Multiple Y axes
When you have more than one series in a plot, you can select to display more than one
y-axis and linking it to one of the series.
To display multiple y-axes:
1. Click the Styles node and click the series you want on the second y-axis.
2. Click the Y Axis tab.
3. In the Attach to Y Axis box, select Axis Y2.
Steps 1 - 3 can be repeated for more series.

Contour Levels
For the Response Contour and Response Surface plots the Contour Levels page is
available in Format Plot.
In the Contour Levels page you can:
- Change the number of Contour levels.
- Increase/decrease the range by changing the Min and Max values.
- Change the coloring scheme in the Begin and End Color boxes.
- Change the colors of individual levels by clicking a level and selecting a new
color in the color box to the right of the Individual level color section.
- Remove a level by marking it and clicking the Remove button.
- Add a specific level by typing a value in the field under the Add-button and
then clicking Add.
In the Contour Level Line Style page you can:
- Select to not display the contour lines by clearing the Show lines check box.
- Change the Pattern, Width and Color of the contour lines by introducing
changes in the respective boxes.
Plot and list contextual tabs
471

See also the Response Contour plot options subsection in the Plot/List chapter.
Multi plots
In SIMCA there are two multi plots: Control Charts and Wavelet Structure.
For these two plots there are Plot nodes beneath the Styles node, named Plot 1, Plot 2,
etc.
In these plot specific nodes the Sub Title and Plot area of the individual plots can be
customized as described in Titles and Background respectively.
Show group
The Show group contains a number of features that can be turned on and off for the
active plot.
- Maximize Plot Area - see the Maximizing the plot area later in this
section.
- The Header, Footer, Legend, Axis Titles, Axes, and Timestamps check
boxes which are displayed if selected and hidden when cleared.
- Regression Line - see the Regression line subsection later in this section.
- Resolve Coefficients - see the Coefficient Plot for hierarchical top level
models subsection in Diagnostics & Interpretation section in the Home
chapter.

Maximizing the plot area
Each plot displays a plot area and also a header and a footer.
To view only the plot area and not the header and footer, on the Layout tab, click
Maximize Plot Area.
User Guide to SIMCA
472
To again view header and footer, click Maximize Plot Area again.

Regression line
The regression line and equation can be displayed for any 2D scatter or line plot in
SIMCA.
Open a plot, for instance the Observed vs. predicted plot, and select the Regression
Line check box.
The attributes of the line and equation can be changed in Format Plot, in the pages
Region Style, Font, and Label Style.

Marked Items tab
The Marked Items tab is automatically activated when marking in a plot. The groups
available on the Marked Items tab are:
- Create from Marked Items
- Drill Down
- Modify Model
- Layout

Create from Marked Items group
Display the selected items in another plot type or a list by clicking the desired plot/list
icon in the Create from Selection group. The plots and list available here can also be
created by clicking the Create command on the shortcut menu.
Create from Selection group
- Scatter - Creates a scatter plot of the currently selected items.
- Line - Creates a line plot of the currently selected items.
- Column - Creates a column plot of the currently selected items.
Plot and list contextual tabs
473
- List - described in the Creating lists subsection in this section.
Creating plot from selection
To generate a new plot from marked points in the open plot, list, or spreadsheet:
1. Mark
- the points in a plot or
- rows or columns in a list or spreadsheet.
2. On the Marked Items tab, click the desired plot type (Scatter, Line,
Column) in the Create from Marked Items group.
Creating special plots from selection
Extended functionality of the Create from Marked Items group is available from the
shortcut menu and described here.
To generate a new plot from the open plot, list, or spreadsheet:
1. Mark
- the points in a plot or
- cells, rows, or columns in a list or spreadsheet.
- nothing which is the same as marking all.
2. Right-click the plot, list, or selection and then click Create | Plot.
3. In the Create Plot dialog, click the desired plot type. The plot types
available are: Scatter Plot, Line Plot, Column Plot, Scatter 3D Plot,
Normal Probability Plot, Histogram Plot, Control Chart, Wavelet
Structure, Wavelet Power Spectrum, and Dendrogram.
4. To use the window of the current plot for the new plot, select the Reuse plot
window check box.
5. Click OK to create the plot.

Creating list from selection
To generate a new list from marked points in the open plot, list, or spreadsheet:
1. Mark
- the points in a plot or
- rows or columns in a list or spreadsheet.
User Guide to SIMCA
474
2. On the Marked Items tab, in the Create from Marked Items group, click
List.
Creating list extended
Extended functionality of the List feature is available from the shortcut menu and
described here.
To create a list from the open plot, list, or spreadsheet:
1. Mark
a. the points in a plot or
b. cells, rows, or columns in a list or spreadsheet or
c. nothing which is the same as marking all.
2. Right-click the plot, list, or selection and then click Create | List.
When only a cell is marked in a spreadsheet, the following message is
displayed. Select the vectors to be displayed in the list by clicking Variable
or Observation.

When selecting Create List with a dendrogram plot active, a special list is created. For
more, see the Create List for the dendrogram subsection in the Tools tab section
previously in this chapter.
Drill Down group
Create comparison (contribution) plots and observations/variable line plots using the
selected items by clicking a button in the Drill down group. The plots can also be
created by clicking the Create command on the shortcut menu.
The possible plots are:
- Plot XObs and Plot YObs - described in the XObs and YObs line plots
subsection in this section.
- Variable Trend Plot - available when a variable is marked in a plot or list,
for instance in the Loading Column Plot. For BEM this plot is the Variable
BCC for the selected variable.
- Contribution Plot - available after marking an observation. For more see the
Creating plots from plots section in the Plot and list contextual tabs
chapter.
- Combined Contribution - available after marking more than one column in
a batch level contribution plot. For more see the Combined Contribution
plot in batch level models subsection in the Contribution plots section in
the Analyze chapter.

Plot and list contextual tabs
475
Creating XObs and YObs line plots from selection
Plot XObs and Plot YObs are available from the Marked Items tab.
To create XObs for the entire spectra, click Spectra on the Data tab.
To create a line plot of all or a selection of the observations:
1. Mark the observations.
2. Click Plot XObs or Plot YObs on the Marked Items tab. Plot YObs is only
available when there are y-variables defined in the dataset spreadsheet.
Define y-variables either at import or by clicking the Save As Default
Workset button in the Workset dialog, tab Variables, after specifying the y-
variables.
Note: The first secondary variable ID is by default displayed on the x-axis
when it is numerical.

Note: With spectral data, it is particularly useful to display the XObs plot of
all observations.
See also the Color subsection in the Tools chapter for details about coloring the line
plot.
Drill down contribution plots
When a time point or batch point deviates from the expected, a contribution plot
displays which variables that have contributed to the deviation.
In the table, find a description on how to create the different types of contribution plots
using the buttons on the Marked Items tab.
The marking is done using Free Mark when not stated otherwise. Table 1 is valid for
regular projects, batch level models, and for batch evolution models when Batch
Marking Mode is NOT selected. Table 2 is valid when using Batch Marking Mode
(only available for batch evolution models).
User Guide to SIMCA
476
Table 1. Contribution plots for all types of projects when Batch Marking Mode is not
the selected tool.
Plot type Action
1 One point
compared to
the average.
Mark a point and click Point to Average Comparison.
2 One point
compared with
another point.
Click one time point and then on the other. Click Point
to Point Comparison.
3 A group of
points
compared to
the average.
Mark all points in the group by circling them or by
holding down the CTRL key and clicking the points.
Then click Group to Average Comparison.
4 One point
compared to a
group of points.
Mark all points in the group by circling them and then
click the single time point (without holding down the
CTRL key). Click Group to Point Comparison.
5 A group of
points
compared to
another group
of points.
Mark all time points in the first group by circling them or
by holding down the CTRL key and clicking the time
points. Release the CTRL key before starting to mark
the next group.
Then mark all time points in the second group by circling
them or by holding down the CTRL key and clicking the
points. Click Group to Group Comparison.
Table 2. Contribution plots for batch evolution models when comparing entire batches
(Batch Marking Mode the selected tool).
Plot type Action
1 One batch in a
BEM compared to
the group
average.
1. Select Batch Marking Mode by clicking the
Select marking tool.
2. Mark a point in the batch and note that all points
of that batch are marked
3. Click Group to Average Comparison.
2 One batch
compared with
another batch.
1. Select Batch Marking Mode.
2. Mark a batch, and then mark another batch.
3. Click Group to Group Comparison.
3 A group of
batches
compared to the
average.
1. Select Batch Marking Mode.
2. Mark at least one point in each batch by circling
them or by holding down the CTRL key and
clicking the points.
3. Click Group to Group Comparison.
Plot and list contextual tabs
477
Plot type Action
4 One batch
compared to a
group of batches.
1. Select Batch Marking Mode.
2. Mark at least one point in each batch by circling
them or by holding down the CTRL key and
clicking the points.
3. Mark the single batch (without holding down the
CTRL key).
4. Click Group to Group Comparison.
5 A group of
batches
compared to
another group of
batches.
1. Select Batch Marking Mode.
2. Mark at least one point in each batch by circling
them or by holding down the CTRL key and
clicking the points.
3. Release the CTRL key before starting to mark
the next group.
4. Mark at least one point in each batch by circling
them or by holding down the CTRL key and
clicking the points
5. Click Group to Group Comparison.
For more about contribution plots, see the Analyze and Predict chapters.
Drill down plots possible
The Drill down group becomes available after marking one or more points.
The table lists the different buttons possible in the Drill down group, for regular and
BLM. For BLM the group can be a batch or a group of batches.
Marked points in
regular projects
Buttons available after marking
1 One or more observations
in for instance a score plot.
1. One observation
marked.
1. Point to Average Comparison - when
clicked it displays the observation vs.
average contribution plot, where the
average is over all observations in the
workset.
2. A group of
observations marked.
2. Group to Average Comparison - when
clicked it displays the group vs. average
contribution plot.
3. One observation first
marked and then a
group.
3. Point to Group Comparison - when
clicked it displays the observation vs.
group contribution plot.
4. A group first marked
and then another
group.
4. Group to Group Comparison - when
clicked it displays the group vs. group
contribution plot.
2 The above but for
predicted observations
(using the predictionset).
The above but for the predictionset.
User Guide to SIMCA
478
Marked points in
regular projects
Buttons available after marking
3 One or more variables in
any plot, for instance in a
loading or a contribution
plot.
Variable Trend Plot - when clicked displays a
variable plot with one or more series.
4 One or more observations
in any plot.
Plot XObs - when clicked displays a plot with
one or more observation series.
Additionally, there is a number of drill down plots possible for batch evolution models
after marking. These are listed in the table below. This table also lists a couple of
special drill down plots created from BLM.
Marked points in
BEM/BLM
Buttons available after marking
1 BEM. With the Batch Marking
Mode selected, one or more
batches in for instance a score
BCC.
1. One batch marked.
1. Point to Average Comparison -
when clicked it displays the
observation vs. average contribution
plot, where the average is the
average batch.
2. A group of batches
marked.
2. Group to Average Comparison -
when clicked it displays the group vs.
average contribution plot.
3. One batch first marked
and then a group.
3. Point to Group Comparison - when
clicked it displays the observation vs.
group contribution plot.
4. A group of batches first
marked and then another
group.
4. Group to Group Comparison -
when clicked it displays the group vs.
group contribution plot.
2 Time points in BEM (marked
with the Marked Batch Mode
unselected).
1. One observation marked.
1. Point to Average Comparison
2. A group of observations
marked.
2. Group to Average Comparison
3. An observation first
marked and then a
group.
3. Point to Group Comparison
4. A group first marked and
then another group.
4. Group to Group Comparison
3 A variable in a plot from a
BEM.
Variable Trend Plot - displays a variable
plot with control limits. This plot cannot
display more than one variable at a time.
4 A point in a BLM score plot for
a model built on score or raw
variables.
A Sources of Variation, SoV, contribution
plot.
Plot and list contextual tabs
479
Marked points in
BEM/BLM
Buttons available after marking
5 A score (t) contribution column
in a BLM.
A BEM contribution plot.
6 More than one score (t)
contribution column in a BLM.
Combined Contribution - displays a
combined contribution plot.
Modify model
In the Modify Model group, features that modify the unfitted model, or create a new
unfitted modified model, are available.
The available features are described in this section and are listed here:
- Exclude - Excluding marked items
- Include - Including marked items
- Class - Assigning observations to classes
- Class | PLS-DA Model/OPLS-DA Model - described in the Creating a
PLS-DA or OPLS-DA model from plot marking subsection in this
chapter.
- Class | Class Models - described in the Creating class models from plot
marking subsection in this chapter.

Excluding marked items
Exclude observations, variables, expanded terms, lags, and batches, using the Exclude
tool.

Excluding observations or variables
To exclude observations, variables, expanded terms, or lags:
1. Open a plot, scores, loadings, VIP, etc.
2. Mark the items to exclude.
3. Click Exclude.
4. SIMCA builds a new workset, with the selected items excluded. The new
unfitted model becomes the active model.
5. Repeat this operation as many times as needed; all exclusions are done in the
unfitted model.
User Guide to SIMCA
480
Note: The variables or observations are removed only from the new workset
and not from any of the plots.
For how to exclude items using the Variable, Observation, or Marked Items
dockable windows, see the Show/Hide section in the View chapter.
Excluding batches in the batch evolution model
To exclude batches in a BEM:
1. Open a plot displaying the batch.
2. On the Tools tab, in the Highlight group, click Select | Batch Marking
Mode.
3. Click a point in the batch and note that all points in that batch are marked.
4. Click Exclude.
Excluding batches in batch level model creating new batch
evolution and batch level model
When marking batches in the BLM, clicking the arrow under the Exclude tool displays
the options:
1. Exclude (default) and
2. Create new BEM and BLM without marked batches.
Selecting the second choice leads to the following:
1. SIMCA creates and fits a new BEM after excluding the marked batches.
2. SIMCA creates a new batch level dataset and a new BLM with the same
settings as before.
This is very useful when outliers are found in the batch level model, and one needs to
rebuild the BEM and BLM without the outlying batches.
Excluding in predictionsets
When the displayed observations are those of a predictionset, marking and excluding
them excludes the observations from the predictionset.
Including marked items
The Include button creates a new model including only the marked items when there is
no unfitted model. When there is an unfitted model, Include adds items to the unfitted
model.

Plot and list contextual tabs
481
Creating class or DA-model from selection
Create Class Models, Create OPLS-DA and Create PLS-DA are available when two
(or more groups) are marked in any plot displaying observations. If only one class is
marked, Class is available.

To assign observations to classes:
1. Mark the observations in a plot or list.
2. Click Class, either the button or the arrow, and select the class name or
number if listed, or click Class and enter a new class name or number in the
dialog.
3. Repeat this operation until you have selected all the observations you want in
classes.
4. When you are done editing the active model, verify the model type, mark the
CM and click Autofit.
To create class or DA-models:
1. Mark two groups of observations in a plot.
2. Click the Class arrow, and click Create Class Models, Create OPLS-DA or
Create PLS-DA.
3. Click Autofit to fit the model.
Note: All unassigned observations are excluded.
To create class models or a DA-model from a dendrogram plot:
1. Mark in the dendrogram plot so that the desired groups are displayed in
different colors.
2. On the Marked Items tab, click Class | Create Class Models or Create
OPLS-DA or Create PLS-DA.
User Guide to SIMCA
482
Layout group
The presentation of marked items in a plot can be manipulated by changing the label
type, change the formatting, etc. available from the Layout group on the Marked
Items tab.


Changing/adding labels
With a large numbers of variables or observations in a plot it is useful to be able to
have labels on selected items.
1. Mark interesting points in the plot.
2. Click the Labels arrow on the Marked Items tab and select the desired label
type. All available labels are listed.
Note: If you want only labels on some items and there are labels on all in the
plot, remove them by clicking Labels | No labels on the Tools tab first and
then use step 1 and 2.

Hiding marked/unmarked items
To hide the marked items, without removing them from the model, click Hide.
To hide the unmarked, without removing them from the model, click Hide | Hide
unmarked items.
To show hidden items, click the Show All button.
The above can also be done from the Properties dialog in the Item Selection tab
where hidden items can be selected to be displayed again.
Locking rows or columns
Locking rows and columns is available for all lists and spreadsheets when an entire
row or column is marked in the spreadsheet. When locking rows or columns those
rows or columns are always displayed when scrolling the list.
Plot and list contextual tabs
483
To lock rows or columns:
1. Mark the columns or rows to lock in the list or spreadsheet by clicking the
column or row number.
2. On the Marked Items tab click Lock Columns or Lock Rows.
The locked rows or columns are displayed when scrolling to the right or down, see the
example here.


Marked Values - changing attributes of selected items in a
plot
To change the fonts, symbols, size, etc., of only the marked points in a plot, click
Format Symbol and Format Label respectively in the Layout group on the Marked
Items tab. The Format Plot dialog is automatically opened with Custom under Styles
or Custom Label under Labels default selected.
Changing the properties of the custom series will only apply to the marked items. See
also the Mini-toolbar - format plot section earlier in this chapter.
User Guide to SIMCA
484


485
SIMCA-P+ 12 features - where
to find them
Introduction
With the rearrangement of SIMCA into ribbons, features were moved, some renamed,
others found obsolete. In this chapter the features of SIMCA-P+ 12 are listed with their
position and name in SIMCA 13.
The changes in terminology, listed in the table, were introduced. The list is in
alphabetical order of the SIMCA-P+ 12 terminology.
SIMCA-P+ 12 SIMCA 13
Autocolor Autoformat
Batch level Batch level, BL
Batch level project (model) Batch level model, BLM
BM BEM. BM includes BEM and optionally BLM.
Contour Plot Response Contour
Default plot action
Buttons in the Drill down group on the Marked
Items tab.
Dendrogram HCA
Distance to Model DModX, DModX PS
Filter Summary Dataset Summary
Hotellings T2Range Hotellings T2
Import wizard SIMCA import
Loadings Bi plot Biplot
Model Overview Plot, List Summary of Fit, Summary of Fit List
Next component Add
Observation level Batch evolution, BE
Observation level project
(model)
Batch evolution model, BEM
Primary dataset Dataset
User Guide to SIMCA
486
SIMCA-P+ 12 SIMCA 13
Remove Component Remove
Residuals Normal Probability
Plot
Residuals N-Plot
Secondary dataset Dataset
SIMCA-P+ SIMCA
T Predicted Score PS
Validate Permutations
Variable Importance VIP
Y Predicted Y PS
SIMCA-P+ 12 File menu
The items on the File menu in SIMCA-P+ 12 are arranged according to the table below
in SIMCA 13.
Feature in SIMCA-P+ 12 SIMCA 13
File | New, first page of wizard
allows selecting project type
and turn on/off the audit trail.
File | New Regular Project, File | New Batch
Project.
CTRL+N creates regular, CTRL+SHIFT+N
creates batch.

The type of project is specified when creating the
project. The next time you create a project the
project type used last time is default. Use the
submenu to change project type, i.e. File | New
Regular Project | New Batch Project.
The Audit Trail is turned on in File | Options in
SIMCA import.
Open File | Open
Close File | Close
Close All Projects Obsolete.
Each instance of SIMCA can only have one open
project. SIMCA can be opened many times
simultaneously.
SIMCA-P+ 12 features - where to find them
487
Feature in SIMCA-P+ 12 SIMCA 13
Delete | Project File | Manage | Delete Project

Delete | Dataset On the Data tab, in the Modify Dataset
group, Delete Dataset

Delete | Model On the Home tab, in the Workset group, Delete.

Save Project File | Save
Save Project As File | Save As | Save Project As
Save Plot/List As File | Save As | Save Plot/List As
Import | Secondary Dataset File | Import Dataset.
Datasets are equal in this version. This means
that any dataset can be used to build a model
whether it was imported first or last.
Import | Import and Merge By
Observation ID
File | Import Dataset.
In the Workset dialog, Select Data page, several
datasets can be selected. Merge is done
internally for the calculations.
Generate HTML Report File | Generate Report
The placeholder group Dataset was removed.
The Model group items include all selected
datasets.
Create Batch Level Project On the Batch tab, in the Dataset group, Create
Batch Level.
The batch level datasets, BL DS, are created and
the workset dialog automatically opened. For
more, see the Create BLM subsection.
Create Batch Level
Secondary Datasets
Obsolete.
Specifying the predictionset for the BLM is done
using the BE DS. The BE DS is rearranged to fit
the BLM.
User Guide to SIMCA
488
Feature in SIMCA-P+ 12 SIMCA 13
Switch Batch Level Obsolete. BEM (observation level) and BLM
(batch level) reside in the same project.
Configure Batch Project for
On-Line
Obsolete. Handled by the on-line software.
Print Setup File | Print | Print Setup
Print Preview File | Print | Print Preview
Print File | Print | Print
Encrypt File | Manage | Encrypt
Compact Project File File | Manage | Compact Project File
Recent file list Recent files when clicking File.
The most recently opened projects are listed and
can be pinned.
Exit File tab, button below Recent files, named Exit
SIMCA.

SIMCA-P+ 12 features - where to find them
489
Import wizard
When importing a new dataset, the SIMCA-P+ 12 import wizard buttons and
commands are available in the SIMCA import in SIMCA 13 according to the tables in
the subsections that follow.


User Guide to SIMCA
490
SIMCA-P+ 12 Import wizard spreadsheet - identifiers and
variable roles
In the import wizard spreadsheet of SIMCA-P+ 12 identifiers and variable roles were
specified. These buttons/shortcut menu commands/arrow menu
commands/Commands button commands were arranged according to the table below,
in the SIMCA import in SIMCA 13.
Feature in SIMCA-P+ 12 SIMCA 13
Identifier specification available
from buttons, column/row arrow
menus and from Commands |
Column/Row Formatting
Specification of identifiers is done in the
Identifiers group on the Home tab.


Variable role specification
available from the button in the
Data section, from the column
arrow menu and from
Commands | Column
Formatting

Variable role specification is done on the Home
tab, in the Variable Types group.

To specify a quantitative x-variable, mark the
variable and click Quantitative and X Variables.
To specify a quantitative y-variable, mark the
variable and click Quantitative and Y Variables.
To specify a qualitative x-variable, mark the
variable and click Qualitative and X Variables.
To specify a qualitative y-variable, mark the
variable and click Qualitative and Y Variables.
To specify a variable as x and Date/Time, mark
the variable and click Date/Time and X
Variables.
To specify a variable as y and Date/Time, mark
the variable and click Date/Time and Y
Variables.
SIMCA-P+ 12 features - where to find them
491
Feature in SIMCA-P+ 12 SIMCA 13
Specifying batch conditions,
available from the button in the
Data section, from the column
arrow menu and from
Commands | Column
Formatting.

To specify a quantitative x-variable, mark the
variable and click Quantitative, Batch
Conditions, and X Variables.
To specify a quantitative y-variable, mark the
variable and click Quantitative, Batch
Conditions, and Y Variables.
To specify a qualitative x-variable, mark the
variable and click Qualitative, Batch Conditions,
and X Variables.
To specify a qualitative y-variable, mark the
variable and click Qualitative, Batch Conditions,
and Y Variables

SIMCA-P+ 12 Import wizard spreadsheet additional features
For identifiers and variable roles, see SIMCA-P+ 12 Import wizard spreadsheet -
identifiers and variable roles.
Other features available from buttons, arrow menus and Commands were arranged
according to the table below, in the SIMCA import of SIMCA 13.
Feature in SIMCA-P+ 12 Position in SIMCA 13
Enable Audit trail for the import in
the project type page.


Specify to Log events to the Audit trail as
Yes or No in Options on the File tab.


Data with a row marked.

Include | Include Row in the Exclude group
on the Home tab.

User Guide to SIMCA
492
Feature in SIMCA-P+ 12 Position in SIMCA 13
Exclude - button

Exclude in the Exclude group, on the Home
tab.

Missing Value field

Missing Value field in Options on the File
tab.
Auto-Color Dataset On the Edit tab, in the Formatting group,
Apply Formatting | Autoformat.

Auto-Format Column Widths General auto formatting as in MS Excel was
implemented. Double-click when the cursor
shows auto formatting symbol.
Import Another File File | New Spreadsheet.

You can import arbitrarily many files to the
SIMCA import. These can be imported in
separate datasets or merged if desired.
Save As File | Save As | Save Spreadsheet As
If you want to save all spreadsheets and
continue working in the SIMCA import later,
see the Workspace subsection in the
SIMCA import chapter.
Column Formatting | Exclude
Column
Exclude | Exclude Column in the Exclude
group on the Home tab.


SIMCA-P+ 12 features - where to find them
493
Feature in SIMCA-P+ 12 Position in SIMCA 13
Row Formatting | Data Include | Include Row in the Exclude group
on the Home tab.

Row Formatting | Exclude Row Exclude | Exclude Row in the Exclude
group on the Home tab.

Undo, Cut, Copy, Paste, Find On the Home tab, in the Clipboard and
Editing groups.

Some are also available in the Quick
Access Toolbar.

Insert | Rows
Insert | Columns
Delete
Shift
Clear Contents
On the Edit tab, in the Rows & Columns
group.

Clear Contents is available by clicking
Delete | Clear.
Create Index On the Edit tab, in the Indexes group. The
first button creates an index column, the
second an index row.

Merge Columns Merge Columns in the Merge group on the
Edit tab.

User Guide to SIMCA
494
Feature in SIMCA-P+ 12 Position in SIMCA 13
Zoom out Missing Value Map available on the View
tab.

Zoom out was discontinued but feature
available as Missing Value Map.
Transpose Transpose in the Rows & Columns group
on the Edit tab.
Downsize Downsize in the Rows & Columns group
on the Edit tab.

Exclude Phases In the Batch & Phase pane, mark what to
exclude and click the Exclude button in the
toolbar in the pane.
See the Batch & Phase pane batch part
topic in the SIMCA import chapter.
Search and Exclude Find/Replace in the Editing group, on the
Home tab. CTRL+F. For more, see the Find
subsection in the SIMCA import chapter.

Hide Formatting Buttons Minimize the Ribbon on the Quick Access
Toolbar arrow menu.

SIMCA-P+ 12 Import wizard class page
After specifying Class ID clicking Next in SIMCA-P+ 12 opened the class page. The
features of this page are available in the Class pane of the SIMCA import in SIMCA
13.
Feature in SIMCA-P+ 12
Position in SIMCA 13 Class
pane
Rename
Merge
Rename

SIMCA-P+ 12 features - where to find them
495
Feature in SIMCA-P+ 12
Position in SIMCA 13 Class
pane
Reset
Delete
Undo Delete
Move using arrows
Collapse and expand nodes
Merge

Undo on the Quick Access Toolbar or
CTRL+Z.

Exclude

Include

Arrows

Collapse and expand nodes.

For more see the Class pane subsection in the SIMCA import chapter.
SIMCA-P+ 12 Import wizard phase and batch page
After specifying Batch ID and Phase ID clicking Next in SIMCA-P+ 12 opened the
Phase page, Next again opened the Batch page. The features of these pages are
available in the Batch & Phase pane in the SIMCA import of SIMCA 13.
Feature in SIMCA-P+ 12 Position in SIMCA 13
Phase page
Time/Maturity variable for the selected
phase
Rename
Merge
Reset
Delete
Move using arrows
Collapse and expand nodes
With more than one y-variable defined,
specify time/maturity variable for the
selected phase by clicking Configure.
Also rename a phase by clicking
Configure.

Merge

User Guide to SIMCA
496
Feature in SIMCA-P+ 12 Position in SIMCA 13

Undo on the Quick Access Toolbar or
CTRL+Z.

Exclude

Arrows

Collapse and expand nodes

Batch page
Conditional Delete
Delete
Undo Delete

Conditional Exclude

Exclude

Include

For more, see the Batch & Phase pane batch part and Batch & Phase pane phase
part subsections in the SIMCA import chapter.
SIMCA-P+ 12 features - where to find them
497
SIMCA-P+ 12 Import wizard last page
The last page of the import wizard contained the features listed in the table.
Feature in SIMCA-P+ 12 Position in SIMCA 13
Project name field.

File name field in Save As dialog that
opens when clicking Finish in the
Import group on the Home tab.

Destination folder
Available in the Save As dialog opened
when clicking Finish.
Variable ID Names
Secondary variable ID names can be
changed in the spreadsheet. The
primary variable ID name cannot be
changed; it is always Primary ID.
Observation ID Names
Secondary observation ID names can
be changed in the spreadsheet. The
primary observation ID name cannot be
changed; it is always Primary ID.
Missing Value Map
Missing Value Map in the Missing
Values group on the View tab.

Import Local Centering
View Local Centering

Local Centering | Import/View in the
Modify Dataset group on the Data tab
in SIMCA after completed import.

See the Importing local centering and
the View imported local centering
topics in the Data chapter.
User Guide to SIMCA
498
Feature in SIMCA-P+ 12 Position in SIMCA 13
View Batch Conditions
Batch conditions, when imported with
the batch evolution dataset
(observation level) are rearranged and
positioned in a separate dataset named
'BC - dataset'.
On the Home tab in SIMCA after
completed import, click the Dataset
arrow.
View Import Log
Summary pane

Display by selecting it on the View tab.
Use simple mode workset dialog
Toggle on and off by clicking Use
Simple Mode/Use Advanced Mode
button available bottom right in the
Workset dialog.
SIMCA-P+ 12 Edit menu
The items on the Edit menu (Undo, Cut, Copy, Paste, Select All, and Find) in
SIMCA-P+ 12 are found on the Quick Access Toolbar in SIMCA 13. Delete is
unavailable; use the DELETE key on your keyboard.
SIMCA-P+ 12 View menu
The items on the View menu in SIMCA-P+ 12 were arranged according to the table
below in SIMCA 13.
SIMCA-P+ 12 Position in SIMCA 13
Toolbars See the SIMCA-P+ 12 View menu - Toolbars subsection.
Dockable
windows
On the View tab, in the Show/Hide group.

Dockable
windows |
Favorites, Add
Command to
Favorites feature
Discontinued
SIMCA-P+ 12 features - where to find them
499
SIMCA-P+ 12 Position in SIMCA 13
Dockable
windows | What
If
Discontinued
Project Window
In the Show/Hide group on the View tab, see above.
The Project Window is by default docked and default
displayed below the ribbon but can be dragged to different
positions.
To have the project window as a regular window, clear the
Enable docking for the project window check box in
Customize dialog available on the shortcut menu of the
Project Window.
Model Window In the Show/Hide group on the View tab, see above.
Add to Favorites
Add to Report
In the Add group on the View tab.

Interface Mode
Discontinued. Each instance of SIMCA can only have one open
project. SIMCA can be opened many times simultaneously.
Set Active
Project
Discontinued. Each instance of SIMCA can only have one open
project. SIMCA can be opened many times simultaneously.
Full Screen
In the Show/Hide group on the View tab, see the top of the
table.
Project Options
On the File tab, button below Recent files.
The Interface page was removed. Workset mode can be
changed by clicking Use Simple Mode/Use Advanced Mode
in the Workset dialog. The positioning of tabs feature was
discontinued.

General Options
Renamed SIMCA Options. On the File tab, button below
Recent files, see above.
User Guide to SIMCA
500
SIMCA-P+ 12 Position in SIMCA 13
Save As Global
Options
Project Options, General tab. Click Save as General.

SIMCA-P+ 12 View menu - Toolbars
The buttons available from toolbars in SIMCA-P+ 12 are available according to the
table below in SIMCA 13.
Feature in SIMCA-P+ 12 Position in SIMCA 13
Customize toolbar Discontinued. There are no toolbars in SIMCA 13.
Standard toolbar

New, Open, Save
On the File tab, click New Regular Project or
New Regular Project | New Batch Project.
On the File tab, click Open.
On the File tab, click Save.
Cut, Copy, Paste
Quick Access Toolbar.

Print On the File tab, click Print.
New workset
On the Home tab, in the Workset group, click
New.
Fit buttons, for ex. Autofit,
Calculate the two first
components, Calculate next
component etc.
On the Home tab, in the Fit model group.
The Zero component model command was
discontinued. Click Add and then Remove for a
zero component model.
Shortcut Bar

SIMCA-P+ 12 features - where to find them
501
Feature in SIMCA-P+ 12 Position in SIMCA 13
Active Model
Displayed in the caption of the Project Window.
Switching active model is done by marking a new
model in the Project Window.
The Automatic change of active model can be
turned off by selecting No in Automatically
change active model in the SIMCA Options
dialog, More Options tab.
Validate
On the Analyze tab, in the Validate group, click
Permutations.
Create four overview plots
On the Home tab, in the Diagnostics &
Interpretation group, click Overview.
Plot R2 and Q2
On the Home tab, in the Diagnostics &
Interpretation group, click Summary | X/Y
Overview.
Score Scatter Plot
On the Home tab, in the Diagnostics &
Interpretation group, click Scores.
Loading Scatter Plot
On the Home tab, in the Diagnostics &
Interpretation group, click Loadings.
DModX Plot
On the Home tab, in the Diagnostics &
Interpretation group, click DModX.
Visualization Plots
Discontinued. Add the 4 plots: score scatter plot
and the XVar line plot of the first 3 variables to a
group in Favorites and open together.
Quick Info
On the View tab, in the Show/Hide group, select
the Quick Info check box.
Plot
The contextual tabs Tools and Layout are opened when a plot is opened and Marked
Items when marking. These contextual tabs are grouped in the Plot umbrella. With a
list active, the List umbrella holds Tools and after marking Marked Items.

Select plot items
On the Tools tab, in the Highlight group, click
Select.
The No Mark tool is no longer available.
Insert text Discontinued.
Zoom plot On the Tools tab, in the Zoom group, click Zoom.
Zoom out
On the Tools tab, in the Zoom group, click Zoom
out.
Read coordinate values
Discontinued. The Status Bar displays the plot
coordinates.
Show/Hide Regression line
On the Layout tab, in the Show group, select the
Regression Line check box.
User Guide to SIMCA
502
Feature in SIMCA-P+ 12 Position in SIMCA 13
Rotate plot
See the Rotating 3D scatter plot subsection in
the Diagnostics & Interpretation section in the
Home chapter.
Add alarm zone Discontinued.
Plot default action
Drill Down group buttons on the Marked Items
tab.
Exclude
On the Marked Items tab, in the Modify Model
group, click Exclude.
Exclude | Create New
Observation Level Model and
Batch Level Project without
Marked Batches.
On the Marked Items tab, in the Modify Model
group, click Exclude | Create new BEM and
BLM without marked batches.
Include
On the Marked Items tab, in the Modify Model
group, click Include.
Note that Include does not work as in the
previous version. See the Including marked
items subsection.
Set Class
On the Marked Items tab, in the Modify Model
group, click Set Class.
For more see the Creating class or DA-model
from selection subsection in the Plot and list
contextual tabs chapter.
Selected item manipulations |
Label Type
To add or change labels for all points in the plot,
click Labels in the Properties group on the Tools
tab.
To add or change labels for marked points, click
Labels in the Layout group on the Marked Items
tab.
Selected item manipulations |
Format Marked Items
On the Marked Items tab, in the Layout group,
click Format Symbol or Format Label
depending on what to format.
Selected item manipulations |
Create List
To create a list of for all points in the plot, click
List in the Change Type group on the Tools tab.

To create a list of only the marked points, click
List in the Create from Marked Items group on
the Marked Items tab.
Selected item manipulations |
Hide Marked Items
On the Marked Items tab, in the Layout group,
click Hide.
Selected item manipulations |
Hide Unmarked Items
On the Marked Items tab, in the Layout group,
click Hide | Hide unmarked items.
SIMCA-P+ 12 features - where to find them
503
Feature in SIMCA-P+ 12 Position in SIMCA 13
Property Bar


Model Type, Comp, R2X, R2Y,
Q2.
The data available in the Property bar when a
model window was active in SIMCA-P+ 12 is
available in the Model Window and Project
Window.
Model, Comp, X-Axis, Y-Axis,
Z-Axis, Batch, Observations,
Variables, Y-Variable
On the Tools tab, in the Data group.
Available for plots and lists displaying results from
one model.
Color by
On the Tools tab, in the Properties group, Color
by.
Properties
On the Tools tab, the dialog box launcher in the
Properties group.
Adaptive Predictions Discontinued.
SIMCA-P+ 12 Dataset menu
The items on the Dataset menu in SIMCA-P+ 12 were arranged according to the table
below in SIMCA 13.
Feature in SIMCA-P+ 12 Position in SIMCA 13
Open On the Home tab, click Dataset.
Quick Info | Variables,
Observations
On the View tab, in the Show/Hide group, select
the Quick Info check box.
Trimming Overview
On the Data tab, in the Summary group, click
Trimming Overview.
Generate Variables
On the Data tab, in the Modify Dataset group,
click Generate Variables.
Merge
On the Data tab, in the Modify Dataset group,
click Merge.
Transpose Dataset
On the Data tab, in the Modify Dataset tab, click
Transpose.
Spectral Filters
On the Data tab, in the Filters tab, click Spectral
Filters.
The individual filters Derivatives, MSC, and SNV
are available from the menu.
Time Series Filters
On the Data tab, in the Filters group, click Time
Series Filters.
User Guide to SIMCA
504
Feature in SIMCA-P+ 12 Position in SIMCA 13
Export Dataset Discontinued. Use File | Save As | Save List As.
Filter Summary
On the Data tab, in the Summary group, click
Dataset Summary. In the dialog click the Filter
tab.
Dataset Properties
On the Data tab, in the Summary group, click
Dataset Summary.
SIMCA-P+ 12 Workset menu
The items on the Workset menu in SIMCA-P+ 12 were arranged according to the
table below in SIMCA 13.
Feature in SIMCA-P+ 12 Position in SIMCA 13
Statistics
On the Home tab, in the Workset group, click
Statistics.
New
On the Home tab, in the Workset group, click
New.
New as Model
On the Home tab, in the Workset group, click
New As.
Edit
On the Home tab, in the Workset group, click
Edit.
Delete
On the Home tab, in the Workset group,
click Delete.
Create Hierarchical Batch
Models
On the Batch tab, in the Dataset group, click
Create Hierarchical Batch Models.
Spreadsheet
Discontinued as separate spreadsheet.
Accessible through the Workset dialog,
Spreadsheet tab.
Model Options
On the Home tab, click the Model Options dialog
box launcher in the Workset group.
SIMCA-P+ 12 Analysis menu
The items on the Analysis menu in SIMCA-P+ 12 were arranged according to the table
below in SIMCA 13.
Feature in SIMCA-P+ 12 Position in SIMCA 13
Change Model Type
On the Home tab, in the Workset group, click
Change Model Type.
Change Model Type |
Hierarchical Base Model, Non
Hierarchical Base Model
On the Data tab, in the Base Model group, click
Hierarchical, Non Hierarchical.
SIMCA-P+ 12 features - where to find them
505
Feature in SIMCA-P+ 12 Position in SIMCA 13
Autofit
On the Home tab, in the Fit Model group, click
Autofit.
Next component
On the Home tab, in the Fit Model group, click
Add.
Two First Components
On the Home tab, in the Fit Model group, click
Two First.
Zero Components
Discontinued. Fit model and Remove
components until zero.
Remove Component
On the Home tab, in the Fit Model group, click
Remove.
Autofit Class Models
On the Home tab, in the Fit Model group, click
Autofit with the BEM/CM marked in the Project
Window.
Summary
On the Home tab, in the Diagnostics &
Interpretation group, click Summary of Fit.
Summary | Model Overview
Plot
On the Home tab, in the Diagnostics &
Interpretation group, click Summary of Fit |
Summary of Fit.
The Model Overview Plot and List were
renamed Summary of Fit.
Scores
On the Home tab, in the Diagnostics &
Interpretation group, click Scores.
Loadings
On the Home tab, in the Diagnostics &
Interpretation group, click Loadings.
Loadings Bi plot
On the Analyze tab, in the Analysis group, click
Biplot.
Hotelling's T2Range
On the Home tab, in the Diagnostics &
Interpretation group, click Hotelling's T2.
Coefficients
On the Home tab, in the Diagnostics &
Interpretation group, click Coefficients.
Coefficients | Y Related
Profiles
On the Analyze tab, in the Analysis group, click
Y Related Profiles. Y Related Profiles
Overview was discontinued.
Variable Importance | Plot
On the Home tab, in the Diagnostics &
Interpretation group, click VIP.
Variable Importance | List
Discontinued. To create the list, open the plot and
on the Tools tab, in the Change Type group,
click List.
Residuals | Normal
Probability Plot
On the Analyze tab, in the Analysis group, click
Residuals N-Plot.
User Guide to SIMCA
506
Feature in SIMCA-P+ 12 Position in SIMCA 13
Residuals | List
Not available from button in ribbon. On the
Plot/List tab, in the Standard Plots group, click
List, select YVarRes, YVar, YPred, SerrL,and
SerrU.
Distance to Model
On the Home tab, in the Diagnostics &
Interpretation group, click DModX.
Observed vs. Predicted
On the Home tab, in the Diagnostics &
Interpretation group, click Observed vs.
Predicted.
Contribution
On the Analyze tab, in the Analysis group, click
Contribution.
Observation Risk
Not available from button in ribbon. On the
Plot/List tab, in the Standard Plots group, click
Column and select to plot ORisk.
Batch Control Charts |
Scores Plot
On the Batch tab, in the Analysis Control
Charts, click Scores BCC.
Batch Control Charts |
Distance to Model X Plot
On the Batch tab, in the Analysis Control
Charts, click DModX BCC.
Batch Control Charts |
Variable Plot
On the Batch tab, in the Analysis Control
Charts, click Variable BCC.
Batch Control Charts |
Hotelling's T2Range Plot
On the Batch tab, in the Analysis Control
Charts, click Hotelling's T2 BCC.
Batch Control Charts |
Observed vs. Time/Maturity
Plot
On the Batch tab, in the Analysis Control
Charts, click Obs vs. Time/Maturity BCC.
Batch Control Charts List
On the Batch tab, in the Analysis Control
Charts, click List BCC.
Validate
On the Analyze tab, in the Validate group, click
Permutations.
Batch Variable Importance
On the Batch tab, in the Variable Summary
group, click Variable Importance Plot.
Available when a BLM is the active model.
CV-ANOVA
On the Analyze tab, in the Validate group, click
CV-ANOVA.
SIMCA-P+ 12 Predictions menu
The items on the Predictions menu in SIMCA-P+ 12 were arranged according to the
table below in SIMCA 13.
Feature in SIMCA-P+ 12 Position in SIMCA 13
Specify Predictionset |
Specify
On the Predict tab, in the Specify Predictionset
group, click Specify.
SIMCA-P+ 12 features - where to find them
507
Feature in SIMCA-P+ 12 Position in SIMCA 13
Specify Predictionset |
Specify, Import section.
On the Data tab, in the Modify Dataset group,
click Local Centering.
Local centering does not discriminate
predictionsets.
Specify Predictionset |
Specify, Spreadsheet
Discontinued. Add new observations in a dataset
and select the new observations as predictionset.
Specify Predictionset |
Dataset
On the Predict tab, in the Specify Predictionset
group, click As Dataset.
Specify Predictionset |
Complement
WS/Complement WS Batches
On the Predict tab, in the Specify Predictionset
group, click Complement WS/Complement WS
Batches.
Specify Predictionset | As
Workset
On the Predict tab, in the Specify Predictionset
group, click As Workset.
Specify Predictionset | Class
On the Predict tab, in the Specify Predictionset
group, click Class.
Adaptive Predictions Discontinued.
Save as Secondary Dataset
Discontinued. Predictionsets can be saved under
a specified name without saving a dataset. See
the Specify subsection in the Predict chapter.
Y Predicted
On the Predict tab, in the Plots group, click Y
PS.
T Predicted
On the Predict tab, in the Plots group, click
Score PS.
Hotelling's T2Range
On the Predict tab, in the Plots group, click
Hotelling's T2PS.
Distance to Model
On the Predict tab, in the Plots group, click
DModX PS.
Time Series
On the Predict tab, in the Plots group, click Time
Series PS.
Contribution
On the Predict tab, in the Plots group, click
Contribution PS.
Coomans' Plot
On the Predict tab, in the Classification group,
click Coomans' plot.
Classification List
On the Predict tab, in the Classification group,
click Classification List.
Misclassification Table
On the Predict tab, in the Classification group,
click Misclassification Table.
Prediction List
On the Predict tab, in the List group, click
Prediction List.
Control Charts
On the Predict tab, in the Plots group, click
Control Charts PS.
User Guide to SIMCA
508
Feature in SIMCA-P+ 12 Position in SIMCA 13
Batch Control Charts |
Scores Plot
On the Batch tab, in the Prediction Control
Charts, click Scores PS BCC.
Batch Control Charts |
Distance to Model X Plot
On the Batch tab, in the Prediction Control
Charts, click DModX PS BCC.
Batch Control Charts |
Variable Plot
On the Batch tab, in the Prediction Control
Charts, click Variable PS BCC.
Batch Control Charts |
Hotelling's T2Range Plot
On the Batch tab, in the Prediction Control
Charts, click Hotelling's T2 PS BCC.
Batch Control Charts |
Observed vs. Time/Maturity
Plot
On the Batch tab, in the Prediction Control
Charts, click Obs vs. Time/Maturity PS BCC.
Batch Control Charts List
On the Batch tab, in the Prediction Control
Charts, click List PS BCC.
SIMCA-P+ 12 Plot/List menu
The items on the Plot/List menu in SIMCA-P+ 12 were arranged according to the
table below in SIMCA 13.
Feature in SIMCA-P+ 12 Position in SIMCA 13
Scatter Plots
On the Plot/List tab, in the Standard Plots
group, click Scatter.
Scatter 3D Plot
On the Plot/List tab, in the Standard Plots
group, click Scatter 3D.
Line Plots
On the Plot/List tab, in the Standard Plots
group, click Line.
Time Series Plots
On the Plot/List tab, in the Control Charts
group, click Time Series.
Column Plots
On the Plot/List tab, in the Standard Plots
group, click, Column.
Histogram Plots
On the Plot/List tab, in the Custom Plots group,
click Histogram.
Contour Plots
On the Plot/List tab, in the Custom Plots group,
click Response Contour.
Response Surface Plots
On the Plot/List tab, in the Custom Plots group,
click Response Surface.
Normal Probability Plots
On the Plot/List tab, in the Custom Plots group,
click Normal Probability.
Dendrogram
On the Analyze tab, in the Clustering group,
click HCA.
Batch Plots | Observed vs.
Smoothed Y Plot
On the Batch tab, in the Time/Maturity group,
click Observed vs. Smoothed Y.
SIMCA-P+ 12 features - where to find them
509
Feature in SIMCA-P+ 12 Position in SIMCA 13
Batch Plots | Unaligned vs.
Aligned Plot.
On the Batch tab, in the Time/Maturity group,
click Unaligned vs. Aligned Plot.
Control Charts
On the Plot/List tab, in the Control Charts
group, click Control Charts.
Wavelet Structure
On the Plot/List tab, in the Custom Plots group,
click Wavelet Structure.
Wavelet Power Spectrum
On the Plot/List tab, in the Custom Plots group,
click Wavelet Power Spectrum.
Lists
On the Plot/List tab, in the Standard Plots
group, click List.
Export to Spotfire Discontinued.
Step Response Plot
On the Plot/List tab, in the Custom Plots group,
click Step Response Plot.
Filter Summary
On the Data tab, in the Summary group, click
Dataset Summary. The Filter tab is available
after filtering the dataset.
Visualization plots
Discontinued. Add the 4 plots: score scatter plot
and the XVar line plot of the first 3 variables to a
group in Favorites and open together.

SIMCA-P+ 12 Window menu
The items on the Window menu in SIMCA-P+ 12 were positioned in the Window
group on the View tab in SIMCA 13, according to the table below.
Feature in SIMCA-P+ 12 Position in SIMCA 13
Close Click the cross in the caption of the window.
Close All
On the View tab, in the Window group, click
Close | Close All.
Cascade
On the View tab, in the Window group, click
Cascade.
Tile Horizontally
On the View tab, in the Window group, click Tile
Horizontally.
Tile Vertically
On the View tab, in the Window group, click Tile
Vertically.
Open items list Discontinued.
Windows
On the View tab, in the Window group, click the
Window Options dialog box launcher.
User Guide to SIMCA
510
SIMCA-P+ 12 Help menu
The commands on the Help menu in SIMCA-P+ 12 are found on the ?-mark menu in
SIMCA 13.

SIMCA-P+ 12 shortcut menu
The commands on the context sensitive shortcut menu for plots and lists in SIMCA-P+
12 were arranged according to the table below in SIMCA 13. The items are generally
also still available from the shortcut menu but that is only mentioned in this table when
it is the only place the item is available from.
Feature in SIMCA-P+ 12 Position in SIMCA 13
Add Observations, Add
Variables
Same.
Show as Histogram
The quick shortcut was discontinued. Create |
Plot, Histogram on the shortcut menu or view the
histogram in the Quick Info.
Create
On the Tools tab, in the Change Type group.
With marking, on the Marked Items tab, in the
Create from Marked Items and the Drill Down
groups.
For more, the SIMCA-P+ 12 shortcut menu -
Create subsection.
Add to Favorites
On the View tab, in the Add group, click Add to
Favorites.

Add to Report
On the View tab, in the Add group, click Add to
Favorites.
Format Date/Time Same.
Lock the First xx Rows and
Lock the First xx Columns.
On the Marked Items tab, in the Layout group,
click Lock...
Sort Ascending and Sort
Descending
On the Tools tab, in the Highlight group, click
Sort | Sort Ascending or Sort Descending.
Maximize Plot Area
On the Layout tab, in the Show group, click
Maximize Plot Area.
Plot Settings
Renamed Format Plot and is available on the
Layout tab, to the far right.
Properties
Same with some exceptions. See the SIMCA-P+
12 Properties dialog subsection.
Quick Info
On the View tab, in the Show/Hide group, select
the Quick Info check box.
Reset rotation Same.
SIMCA-P+ 12 features - where to find them
511
Feature in SIMCA-P+ 12 Position in SIMCA 13
Save Plot or List As On the File tab, Save As.
Save Plot Settings | Default
On the Layout tab, in the Templates group, click
Save as default.
On the shortcut menu, renamed Plot Templates |
Save as default.
Save Plot Settings | saved
plot settings
On the Layout tab, in the Templates group, click
Save as.
On the shortcut menu, renamed Plot Templates |
Save as and select the old plot formatting file in
the dialog.
Save Plot Settings |
Configurations which allows
saving to new name and to
delete custom configurations.
To save to a new template: On the Layout tab, in
the Templates group, click Save as. On the
shortcut menu, renamed Plot Templates | Save
as and select the old plot formatting file in the
dialog.
To delete a template: On the Layout tab, in the
Templates group, click Open templates folder
and delete the unwanted file.
On the shortcut menu, (renamed) Plot Templates
| Save as/Open templates folder.
Switch Plot Settings | Default
On the Layout tab, in the Templates group, click
Load.
On the shortcut menu, renamed Plot Templates |
Save as default.
Switch Plot Settings | saved
plot settings
On the Layout tab, in the Templates group, click
Load | saved template.
On the shortcut menu, renamed Plot Templates |
Load | saved template and select the plot
formatting file in the dialog.
Switch Plot Settings |
Configurations
On the Layout tab, in the Templates group, click
Load/Save as/Open templates folder depending
on desired action.
On the shortcut menu renamed Plot Templates |
Load/Save as/Open templates folder.
Zoom Out On the Tools tab, in the Zoom group.

User Guide to SIMCA
512
SIMCA-P+ 12 shortcut menu - Create
The shortcut menu item Create commands SIMCA-P+ 12 were arranged according to
the table below in SIMCA 13.
Feature in SIMCA-P+ 12 Position in SIMCA 13
List
On the Tools tab, in the Change Type group,
click List.
With marking: On the Marked Items tab, in the
Create from Marked Items, click List.
Plot
On the Tools tab, in the Change Type group,
click the desired plot type or click the dialog box
launcher.
With marking: On the Marked Items tab, in the
Create from Marked Items, click the desired
plot type or click the dialog box launcher.
Specifically for the Sources of Variation plot,
click Column to display batch conditions too.
Plot XObs, Plot YObs
On the Marked Items tab, in the Drill Down
group, click Plot XObs, Plot YObs.
Variable Line Plot
VarDS Line Plot
On the Marked Items tab, in the Drill Down
group, click Variable Trend Plot.
Contribution Plot
On the Marked Items tab, in the Drill Down
group, click the comparison plots.
Combined Contribution
On the Marked Items tab, in the Drill Down
group, click Combined Contribution.
Merge List for HCA
dendrogram.
On the Tools tab, in the Change Type group,
click Merge List.
Variable Batch Control Chart
after marking a variable in a
BEM variable plot.
On the Marked Items tab, in the Create from
Marked Items group, click Variable Trend Plot.
PLS-DA Model after marking
two or more observation
groups.
On the Marked Items tab, in the Modify Model
group, click Class | Create PLS-DA Model.
Class Models after marking in
an observation plot.
On the Marked Items tab, in the Modify Model
group, click Class | Class...
Sources of Variation plot for a
Loading or Contribution plot
from a model built from raw or
score variables.
On the Tools tab, in the Change Type group,
click Sources of Variation.
Toggle between the Sources of Variation plot
and the Column plot by clicking the respective
button in the Change Type group.
Out of Control Summary Plot
(OOC) for BCC plots.
On the Tools tab, in the Change Type group,
click Out of Control Summary.
Batch Control Chart when the
Out of Control Summary plot
is the active plot.
On the Tools tab, in the Change Type group,
click Batch Control Chart.
SIMCA-P+ 12 features - where to find them
513
SIMCA-P+ 12 Color in Properties dialog
This table describes the coloring types in the Color page that work differently in
SIMCA 13.
A general change is that the colors of the categories and continuous cannot be changed
in the Properties dialog in SIMCA 13. Colors can instead be changed and saved in
Format Plot.
Color by SIMCA-P+ 12 SIMCA 13
Batch
Conditions
Color by Batch
Conditions available
after importing batch
conditions with the
observation level dataset.
Color by batch conditions is
available as color by Categories or
Continuous after selecting a batch
condition dataset.
Hierarchical
Variables
For hierarchical top
models the variables can
be colored according to
base model by selecting
by Hierarchical
Variables.
Coloring by hierarchical variables is
done by selecting the hierarchical
dataset in the Data box and then
selecting the variable. Alternatively
by selecting the model in the Data
box, XVar in the Item box and the
hierarchical variable in the X
Variables box.
Score
Categories
Available for any plot
displaying observations.
Color by Score Categories is a
subset of coloring by Vector
(categories).
What If Colors the predictionset
observations originating
from the What If in a
different color from the
original predictions.
The What If was discontinued and
consequently this coloring type.

User Guide to SIMCA
514
SIMCA-P+ 12 Properties dialog
The following property pages are no longer available in the Dataset Properties dialog:
- Batch Conditions are found as batch level datasets, prefixed BC when
automatically created at the import of the batch evolution (observation level)
dataset.
- Local centering is found by clicking Local centering | View in the Modify
Dataset group on the Data tab.
- Default centering - Default centering used for predictionsets was
discontinued. Predictionsets use local centering if available. How local
centering is done when unavailable for the predictionset is described in the
Local centering missing subsection.
The Properties dialog for the Response Contour and Response Surface plots no
longer include the following features:
- Use color for the Response Contour plot. To display the plot without
colors, change the Begin and End colors to white in the Contour Levels
page in the Format Plot dialog.
- Rotation of axis labels, previously available in the Labels page, is now
available in the Format Plot dialog, in the Axis node.
- Save as Default Options, previously available in the Labels page, is now
available in the Project Options dialog, Plot Labels page only.

SIMCA-P+ 12 Plot Settings
The 2D plot library was rewritten resulting in many changes in the plot settings pages.
The three pages Axis, Header and Footer, and Plot Area are combined in the new
Format Plot dialog where the:
- Axis features are found in the Axis node.
- Header and Footer features are found in the Titles node.
- Plot Area features are found in the other nodes where Background, Legend
and Limits mainly contain the same features.
SIMCA-P+ 12 features - where to find them
515
This section describes where to find old functionality. See the Format Plot section to
read about all features in the Format Plot dialog.
Features with the same name positioned in the same place are not listed here.
In SIMCA-P+ 12 Point of the feature
Placement in
SIMCA 13
Axis
Increment by (major
ticks)
To define spacing
between major tick
marks.
Step size in the
individual axis pages.
Minor tick spacing
To define spacing
between minor tick
marks.
Discontinued.
Decimal places
To define how many
decimals to display in the
axis annotation.
Discontinued.
Recalculate Scale
Resets minimum and
maximum to default.
Renamed Always
recalculate scales.
Tick Mark Label -
Scientific, Time
To display the scale in the
specified unit.
Renamed Annotation
and discontinued except
for Rotation.
Rotation
Rotating the tick mark
label/annotation.
In the Annotation
section in the respective
Axis pages.
Logarithmic scale
To display the axis in
logarithmic scale.
Discontinued.
Values in reverse order
To display the axis with
the highest number to the
left and lowest to the
right.
Reverse axis in the
respective Axis pages.
Always recalculate
scales
When cleared the plot
was not updated with a
new value range.
Clearing the Autoscale
check box in the Axis
General page results in
that the plot is not
updated with new values
when switching vectors
one the axes. An
exception is when the
range of the new vector is
larger; in that case the
scale is updated to show
all values.
Aspect ratio available
when clicking Aspect
ratio under Axes.
To define the ratio
between the x and y-axis.
Discontinued.
Font Effects Strikeout
and Underline
To display the annotation
or axis title underlined or
with a line through the
text.
Discontinued.
User Guide to SIMCA
516
In SIMCA-P+ 12 Point of the feature
Placement in
SIMCA 13
Header and Footer
Border style
To customize the line
style of the border around
the header or footer.
Discontinued.
Font Effects Strikeout
and Underline
To display the header or
footer underlined or with a
line through the text.
Discontinued.
Plot Area where Series was renamed Styles
Show symbols
To show/hide symbols in
scatter and line plots.
In line plots, select a
symbol type in the Styles
node, Symbol Style
page, Shape box (default
None for line plots). In
scatter plots you can
select another shape but
not None.
Show line
To show the line between
points in a scatter plot.
For scatter plots there is
no Line Style page which
means that this feature
was discontinued.
However, if you create
the line plot, you can
select to display symbols
by selecting a Shape in
the Symbol Style page
in the Styles node.
Show data labels To hide the data labels.
Handled by the
Properties dialog.
Anchor and Alignment
of Data Labels
To align the plot label.
In the Label Style page
in the Labels node,
Position specifies where
to put the label with
respect to the point.
Offset of data labels
To specify how far from
its item the label should
be placed.
Labels Style page in the
Labels node.
Use the symbol color
To display the label in the
same color as the point.
The labels are by default
displayed in the same
color as the outline of the
symbols. By default the
outline of the symbols is
dark gray. To display the
labels in the same color
as the interior of the
symbol, in the Symbol
Style page, change the
Outline color to
Automatic.
SIMCA-P+ 12 features - where to find them
517
In SIMCA-P+ 12 Point of the feature
Placement in
SIMCA 13
Font Effects Strikeout
and Underline
To display the legend or
plot labels underlined or
with a line through the
text.
Discontinued.
Background
To customize the
background of the plot.
The functionality has
been changed for this
version allowing gradient
fill but discontinuing
border styles.
Is boxed was
discontinued.
Ellipse
Ellipse page Set the
ellipse coordinates: To
distort the ellipse.
The Line Style features
for the ellipse are found in
the Limits and
Regions node.
Discontinued to avoid
distortion of the ellipse.
Alarm Zones
To mark an area of
interest.
Discontinued.
Location
To change the positioning
of the plot area in the plot
window.
Discontinued.
Plot Area Margins
To change the margins
around the plot area in
reference to the plot
window.
Discontinued.
Rotation of data labels
To specify the direction of
the data label.
Labels Style page in the
Labels node.
Alignment of data
labels
To align somehow. Only
when the label is
displayed on more than
one line.
Discontinued.
Text color
To color all labels in one
specific color.
In the Labels node,
Labels Style page,
Label section, select the
desired color in the Color
box.
Anchor
To specify where to
display the item.
Position in the Label
Style page in the Labels
node.
Column Style Pattern
To display a pattern on
the columns.
Gradient Fill, in Column
Style in the Styles node,
replaces this feature.
Overlap
Enabling overlapping of
columns
Discontinued.
User Guide to SIMCA
518
In SIMCA-P+ 12 Point of the feature
Placement in
SIMCA 13
Tick Marks in 3D scatter
plots
Customizing how the ticks
are displayed.
Click Axis General and
Tick Marks.
Marked values
To change symbol style,
symbol color, label color,
or font, marking style.
The Custom style is
created after marking and
clicking Format Symbol
or Format Label. The
marked item formatting
can be changed here.
Bubble-page available
after sizing score vectors.
Specifying the size of the
bubbles in Min and Max
fields.
Styles node, Symbol
Style, vector series, Size
field. Here you cannot
specify minimum and
maximums, just enlarge
or shrink all symbols
proportionally.
Limit Zero in Control
Charts
To customize the zero-
line.
Discontinued.
Display the Surface
Mesh in Contour Levels
To display the Contour
Plot as a surface plot.
Discontinued. Create the
surface plot by clicking
the Response Surface
button.
Display the Contour
Level lines in Contour
Levels
To display/hide the
contour lines.
Select/clear the Show
lines check box in the
Contour Level Line
Style page in the
Contour node.
3D Label - Contour Plot
To customize the label
size, font etc.
See the Labels
subsection in the Layout
tab section in the Plot
and list contextual tabs
chapter.
View 3D To customize the rotation.
Turn the plot as desired
by holding down the
mouse-button and
moving the mouse.
Managing plot settings
Delete plot settings
Being able to delete
unwanted plot settings
configurations.
Plot Settings is now
called Format Plot.
Different configurations
can be saved to file.
These files are handled
as any other files.
On the Layout tab, in the
Templates group,
clicking Open templates
folder enables deleting
unwanted templates.
SIMCA-P+ 12 features - where to find them
519
In SIMCA-P+ 12 Point of the feature
Placement in
SIMCA 13
Managing plot settings
Saving/replacing/restoring
plot settings.
The managing of plot
formatting is done in the
Templates group on the
Layout tab.


521
Help
Introduction
This chapter describes all commands on the Help menu.
Use the Help menu to: access the help, register your product, access the Umetrics web
site, or view version details.
The SIMCA help is based on the user guide. The user guide documents are transferred
to a compiled HTML file. To read the Help file, Internet Explorer version 6.0 or higher
must be installed but does not need to be the default browser.
HTML help
The HTML help file is installed to include interactive and stand alone help.
Open the help by:
- Clicking Contents/Search/Index from the help menu .
- Clicking the Help-button in one of the dialogs or wizards.
- Clicking SIMCA Help from the Program menu.
Use the Contents, Index, or Search tabs to find what you are looking for.
Note: Using citation marks <"> allows searching of phrases.
Additionally, the Analysis Advisor is available to guide you through the analysis.
Open it by selecting the Analysis Advisor check box in the Show/Hide group on the
View tab. For more, see the Analysis Advisor subsection in the View chapter.
Register
After installing SIMCA, if your license needs an activation key, SIMCA will issue a
message requesting you to contact Umetrics to register.
If you choose to register later, from the Help menu, click Register and follow the
directions.
Umetrics on the Web
If you have an Internet connection, you can visit the web page of Umetrics
(www.umetrics.com) to get the latest news and other information by clicking
Umetrics on the Web on the Help menu.
User Guide to SIMCA
522
About SIMCA
To find the version number of SIMCA, click About SIMCA on the Help menu.

523
Appendix A: Statistics
Introduction
This chapter contains brief background information and formulas for SIMCA. For
more extensive reading, use the reference material.
The topics covered are:
- PCA - Principal Components modeling
- PLS - Partial Least Squares Projection to Latent Structures modeling
- OPLS/O2PLS - Orthogonal PLS modeling
- Cluster Analysis (CA), dendrograms, Hierarchical CA (HCA), PLS
including Hierarchical Cluster Analysis - HCA and PLS-Trees
- Vectors available in SIMCA
- Formulas and descriptions
- Transform page criteria
- Scaling
- Cross validation
- PLS Time Series Analysis
- CV-ANOVA
- Fisher's Exact test
- Control Chart statistics
- S-plot
Fit methods background
A dataset is composed of N rows and K columns. The N rows are here called
observations and the K columns are called variables.
Geometrically we can represent the observations as points in a multidimensional space
where the variables define the axes. The lengths of the axes are determined by the
scaling of the variables.
User Guide to SIMCA
524
To calculate a model that approximates the dataset the following fit methods are
available in SIMCA:
- Principal Components modeling - PC modeling.
- Partial Least Squares Projection to Latent Structures modeling - PLS
modeling.
- Orthogonal PLS - OPLS and O2PLS modeling.
All other fit methods, available by clicking Change Model Type use one of the fit
methods above, see the Model types available subsection in the Home chapter.
PCA - Principal Components modeling
Principal Component Analysis is the technique for finding a transformation that
transforms an original set of correlated variables to a new set of uncorrelated variables,
called principal components. The components are obtained in order of decreasing
importance, the aim being to reduce the dimensionally of the data.
The analysis can also be seen as an attempt to uncover approximate linear
dependencies among variables.
PC modeling shows the correlation structure of your data matrix X, approximating it
by a matrix product of lower dimension (TP'), called the principal components plus a
matrix of residuals (E).
X = Xbar + TP + E
where
Xbar contains X average.
T is a matrix of scores that summarizes the X-variables.
P is a matrix of loadings showing the influence of the variables.
E is a matrix of residuals; the deviations between the original values and the
projections.
This geometrically corresponds to fitting a line, plane or hyper plane to the data in the
multidimensional space with the variables as axes. The scaling of the variables
specifies the length of the axes of this space. For more about scaling see the Scaling
section later in this chapter.
SIMCA iteratively computes one principal component at a time, comprising a score
vector t
a
and a loading vector p
a
.
Number of model dimensions (A)
To get an overview of the dataset, a few (2 or 3) principal components are often
sufficient.
However, if the PC model is used for modeling or for other predictions (e.g. principal
properties), cross validation (CV) should be used for testing the significance of the
principal components. In this way, the significant number of PC components, A, is
obtained, which is essential in modeling.
It is also possible to use the eigenvalue limit (EV) for testing component significance
by clearing the Use cross validation when fitting check box in Model Options, tab
Model.
Eigenvalue Limit (EV): A component is considered significant if its normalized
eigenvalue is larger than 2.
Appendix A: Statistics
525
Cross validation (CV) is described in the Cross validation section later in this
chapter.
Reference PCA
1. Jackson, J.E. (1991), A Users Guide to Principal Components, John Wiley,
New York. ISBN 0-471-62267-2.
PLS - Partial Least Squares Projection to Latent
Structures modeling
When fitting a Partial Least Squares Projection to Latent Structures model, PLS finds
the linear (or polynomial) relationship between a matrix Y (dependent variables)
and a matrix X (predictor variables) expressed as:
Y = f(X) + E
The matrix X refers to the predictor variables and their squared and/or cross terms if
these have been added. Active X variables or expansions participating in the model are
sometimes referred to as terms. The function f(X) is usually a polynomial, possibly in
the transformed variables.
Note: Cross terms should be added to the linear terms in X only when the
data supports such terms, for instance after using a design to generate X.
PLS is most easily understood geometrically, where we see the matrices X and Y as N
points in two spaces, the X-space with K axes, and the Y-space with M axes, K and M
being the number of columns in X and Y.
PLS modeling consists of simultaneous projections of both the X and Y spaces on low
dimensional hyper planes. The coordinates of the points on these hyper planes
constitute the elements of the matrices T and U. The analysis has the following
objectives:
- To well approximate the X and Y spaces
- To maximize the correlation between X and Y
The PLS model accomplishing these objectives can be expressed as:
X = Xbar + TP + E
Y = Ybar + UP + F
U = T + H (the inner relation)
where
Xbar contains X average.
Ybar contains Y average.
Note that the coefficients of the inner relation are 1.
In the PLS algorithm there are additional loadings, W, called weights. These express
the correlation between U and X and are used to calculate T.
This modeling geometrically corresponds to fitting a line, plane or hyper plane to both
the X and Y data represented as points in a multidimensional space, with the objective
of well approximating the original data tables X and Y, and maximizing the
covariance between the observation positions on the hyper planes.
User Guide to SIMCA
526
SIMCA will iteratively compute one PLS component at a time, that is: one vector each
of X-scores t, Y-scores u, weights w and c, and loadings p.
The PLS components are calculated in descending order of importance.
Vector Description
T Matrix of scores that summarizes the X variables.
W Matrix of weights expressing the correlation between X and U (Y).
U Matrix of scores that summarizes the Y variables.
C Matrix of weights expressing the correlation between Y and T (X).
E, F, H Matrices of residuals.
Number of model dimensions (A)
The criterion used to determine the model dimensionality, (number of significant PLS
components), is cross validation (CV). With CV, observations are kept out of the
model development, then the response values (Y) for the kept out observations are
predicted by the model, and compared with the actual values.
This procedure is repeated several times until every observation has been kept out once
and only once.
Reference PLS
Wold, S., Sjstrm, M., and Eriksson, L., (2001b), PLS-Regression: A Basic Tool of
Chemometrics, Journal of Chemometrics, 58, 109-130.
OPLS/O2PLS - Orthogonal PLS modeling
OPLS - Orthogonal PLS modeling
A simple way to understand OPLS is to consider how PLS and OPLS differ in their
handling of the variance of the X-matrix. PLS divides the variability in X in two parts
(figure below), i.e., the systematic and residual parts. The systematic part is the sum of
the variability in X that is correlated (predictive) to Y and the variability in X that is
uncorrelated (orthogonal) to Y. Thus, whereas PLS divides the sum of squares of X in
two parts, OPLS divides it in three parts. This yields good predictions and improved
interpretability.
It should be noted that in the single-Y case, by theory, the OPLS model can only have
one predictive component [Trygg and Wold, 2002]. Should a single-Y OPLS model
comprise more than one component, all components beyond the first one reflect
orthogonal variation. However, with multiple Y-variables there can be more than one
predictive OPLS component.
For the single-y case only OPLS is available in SIMCA. The results from
OPLS/O2PLS in SIMCA-P+ 12 are identical to the results using OPLS in SIMCA 13.
Appendix A: Statistics
527

Figure: PLS (left) divides the variability in the X-matrix in two parts, the systematic
variability and the residual variability. OPLS (right) further splits the systematic
variability, R2X, in two parts, the part that is correlated (predictive) to Y and the part
that is uncorrelated (orthogonal) to Y.
In the case of a single response variable, we can write the X-part of the OPLS model as
X = 1x' + tp' + T
o
P
o
' + E
and the OPLS model prediction of y as
y = y' + tq' + f
The multi-Y OPLS model can be expressed as follows:
X = 1x' + TP' + T
o
P
o
'+ E
and the OPLS model prediction of Y as
Y = 1y' + TQ' + F
Where the matrix products TP' and TQ' hold the joint X/Y information overlap. The
number of score vectors (in T and U) and loading vectors (in P' and Q') is determined
using cross-validation.
Thus, in comparison with the single-y OPLS model, the main difference lies in the fact
that there can be more than one predictive component in the multi-y OPLS model. The
number of predictive components is regulated by the number of latent variables in the
information overlap between X and Y, which in turn often is linked to the rank of Y.
O2PLS - Orthogonal PLS modeling
The O2PLS model can be written as follows:
X = 1x' + TP' + ToPo' + E
Y = 1y' + UQ' + UoQo' + F
Where the matrix products TP' and UQ' hold the joint X/Y information overlap. The
number of score vectors (in T and U) and loading vectors (in P' and Q') is determined
using cross-validation. In the example in the Model Window for OPLS and O2PLS
models subsection in the View chapter, there are seven components of this type
(predictive components).
The number of components in the respective set of components is determined using
cross validation.
For any part of the OPLS/O2PLS model, the percentages explained and predicted
variances can be obtained from plots and lists in the software.
For more about the differences between OPLS/O2PLS and PLS, see the Conventional
PLS compared with OPLS and O2PLS subsection.
User Guide to SIMCA
528
Conventional PLS compared with OPLS and O2PLS
PLS
Conventional PLS applies to the two-block (X/Y) regression problem. It uses X to
construct a model of Y, where the objective is to predict the latter from the former for
new samples in the predictionset. In that sense, PLS is unidirectional, i.e., X Y, but
not vice versa.
When X is composed of e.g. spectroscopic data, process readings or measurements
from bio-analytical platforms, there is a risk that systematic variation may reside in X
which is not linearly correlated with Y. Such variability in X is usually called
Orthogonal in X [1]. Although the Orthogonal in X variation can be handled by a PLS
model, it often makes model interpretation more difficult [2].
OPLS
The OPLS method is a recent modification of the PLS method [1-3], which is designed
to handle variation in X that is orthogonal to Y. OPLS separates the systematic
variation in X into two parts, one that is linearly related (and therefore predictive) to Y
and one that is orthogonal to Y. The predictive variation of Y in X is modeled by the
predictive components. The variation in X which is orthogonal to Y is modeled by the
orthogonal components. This partitioning of the X-data provides improved model
transparency and interpretability, and gives very similar predictive power. For OPLS
with one Y the predictive model with the same number of components is identical to
the same PLS model. Similarly to PLS, OPLS is a unidirectional method, where the
scope is the relation X Y.
O2PLS
O2PLS is a generalization of OPLS [4,5]. In contrast to PLS and OPLS, O2PLS is
bidirectional, i.e. X Y. Additionally, with O2PLS it is possible to partition the
systematic variability in X and Y into three parts, (i) the X/Y joint predictive variation,
(ii) the variation in X orthogonal to Y (X-unique Variation), and (iii) the X-unrelated
variation in Y (Y-unique Variation). The variation in X orthogonal to Y can be further
divided into two parts; one part which is equivalent to the component in OPLS
(Orthogonal in X (OPLS)) is a matrix effect that needs to be modeled to achieve the
best possible prediction and interpretation. The second part (Orthogonal in X (PCA))
consists of structured variation that does not affect the prediction but can be interesting
to study to further improve the interpretation of complex multivariate data. The X-
unrelated variation in Y may be modeled in a similar way.

Figure 1. Overview of the O2PLS model relating two data tables to each other. Unique
variation in X also named variation in X orthogonal to Y in the left-hand side of the
Figure. The X/Y joint predictive variation (middle part of the Figure) describes the
predictive variation between X and Y, the information overlap. The Unique variation in
Y also named variation Orthogonal in Y in the right-hand side of the figure.
Appendix A: Statistics
529
The ability to interpret the X/Y joint predictive variation separated from the non-
correlated variation implies that the model interpretation is refined and simplified.
Furthermore, it should be noted that for the single-y case the OPLS and O2PLS
methods are identical. For such a model there can only be one predictive component
expressing the joint X/Y predictive variation [4,5].
Combination with the hierarchical approach
Because of its ability to divide the information in X and Y into different parts, the
O2PLS model is ideally suited for combining with the hierarchical modeling approach.
This is of interest when working with multi-block process data and spectroscopic data
[6,7]. The flexible structure of the O2PLS method also allows setting it up as a filter
for peeling off any undesired systematic variability in your data.
Y-Related Profiles
The Y-Related Profiles are coefficients rotated displaying the pure profiles of the
underlying constituents in X using the assumption of additive Y-variables.
The estimation includes a linear transformation of the coefficient matrix, Bp(Bp
T
Bp)
-1
,
using only the predictive components to compute the coefficients (i.e., the components
orthogonal to Y are not included in the computation of the Y related profile).
References OPLS and O2PLS
1. Trygg, J., and Wold, S., (2002), Orthogonal Projections to Latent Structures
(OPLS), Journal of Chemometrics, 16, 119-128.
2. Trygg, J., (2004), Prediction and Spectral Profile Estimation in Multivariate
Calibration, Journal of Chemometrics, 18, 166-172.
3. Eriksson, L., Johansson, E., Kettaneh-Wold, N., Trygg, J., Wikstrm, M.,
and Wold, S., Multi- and Megavariate Data Analysis, Part II, Method
Extensions and Advanced Applications, Chapter 23, Umetrics Academy,
2005.
4. Trygg, J., (2002), O2-PLS for Qualitative and Quantitative Analysis in
Multivariate Calibration, Journal of Chemometrics, 16, 283-293.
5. Trygg, J., and Wold, S., (2003), O2-PLS, a Two-Block (X-Y) Latent Variable
Regression (LVR) Method With an Integral OSC Filter, Journal of
Chemometrics, 17, 53-64.
6. Gabrielsson, J., Jonsson, H., Airiau, C., Schmidt, B., Escott, R., and Trygg,
J., (2006), The OPLS methodology for analysis of multi-block batch process
data, Journal of Chemometrics, 20, 362-369.
7. Eriksson, L., Dyrby, M., Trygg, J., and Wold, S., (2006), Separating Y-
predictive and Y-orthogonal variation in multi-block spectral data, Journal
of Chemometrics, 20, 352-361.
User Guide to SIMCA
530
Cluster Analysis (CA), dendrograms, Hierarchical CA
(HCA), PLS-Tree
Large datasets are often clustered, grouped, and relationships between variables are
often different in different groups (clusters). Cluster analysis tries to find a natural
grouping (clustering) of a data set so that there is less variation (greater similarity)
within the clusters, and more variation (less similarity) between the clusters. The
difference between on the one hand CA, and on the other hand classification and
discriminant analysis (CDA) is that in the latter the classes are pre-defined by the user,
and each observation in the workset belongs to one of these classes. CA has no
predefined classes, but after the analysis a set of clusters has been found, and each
observation belongs to one of these. In HCA and PT the fineness of the solution, i.e.,
number of clusters and the observation assignment, can be modified by moving the
cursor up and down in the dendrogram (tree diagram).
CA is particularly useful in the analysis of large data sets, often called Data Mining.
The resulting groups, clusters, can lead to the generation of new ideas exploratory
data analysis and often better models for parts of the data.
SIMCA has two approaches of hierarchical clustering.
1. HCA = hierarchical clustering, available on the Analyze tab, in the
Clustering group, by clicking HCA, and
2. PT = PLS-Trees which is a new PLS-based clustering method developed by
Umetrics (PLS-TREE), available on the Analyze tab, in the Clustering
group, by clicking PLS-Tree.
Note that much of what is described regarding dendrograms and coloring applies both
to HCA and PLS-Trees. Hence, the reader is recommended to also read the following
topics and sections:
- Hierarchical Cluster Analysis - HCA later in this section.
- HCA in the Analyze chapter.
- PLS-Trees later in this section.
- PLS-Tree in the Analyze chapter.
Hierarchical Cluster Analysis - HCA
In Hierarchical Clustering Analysis (HCA) a similarity or distance criterion is first
specified by the user (default = Ward, other option: Single Linkage). Then HCA
basically starts with as many clusters as there are observations (N). The two closest
clusters or observation points are merged, thereafter the two closest clusters or points
are again merged, etc., until only one cluster remains. The result is shown by means of
a dendrogram. Open plots, e.g., score plots, are colored according to the clusters
marked in the dendrogram. For more, see the HCA section in the Analyze chapter.
Ward clustering
In Ward clustering the distance measure is a function that measures the error increase
of the model when a pair of clusters are merged into one new cluster. The error
function used in SIMCA is the classical error function which calculates the difference
in the sum of sum of squares around the mean of each cluster before and after merging
two clusters.
This type of clustering works well when the clusters are spherical.
Appendix A: Statistics
531
Single linkage
In single linkage clustering, the distance between a pair of clusters is the Euclidean
distance between the two observations (one in each cluster) closest to each other. This
type of clustering works well when the clusters are warped filaments.
PLS-Trees
PLS-Trees is an approach similar to regression and classification trees [3], but using
the first score vectors of PLS and PLS-DA instead of the original X-variables for
splitting clusters into two.
Thus, the basic idea of a PLS-Tree is to start with a PLS model of the whole data set
(the user must before specify what is X and Y, scaling, etc.), and then split the data
along the sorted 1st score (t) of the PLS model. This split is made to optimize a
criterion with three parts:
a. improving the variance of the score (t) so that when combined, the variance
in the resulting two sub-groups is as small as possible, and
b. analogously for the variance of Y, and
c. do this so that the sizes of the two sub-groups are as equal as possible. Each
of these three criteria is in advance given a relative weight by the user, and
the criterion to be optimized is the combination of the three parts weighted
by the user supplied weights. The user also specifies a minimal size of a
cluster beyond which it cannot be further split.
The weight allows the user to focus on the clustering in the X-space, or in the Y-space,
or in both, and to more or less focus on splitting into groups of approximately the same
size.
Once the split is made and the criterion indicates an improvement, the split goes on
with the two sub-groups, and thereafter with the resulting sub-groups of these, until all
branches have been terminated due to either lack of improvement by further
splitting, or that the clusters are reaching the minimal size.
The result is a hierarchical set of PLS models in a tree structure. Each model is an
ordinary PLS model which can be further modified, investigated, displayed, and
interpreted.
See also the PLS-Tree and HCA - Dendrogram sections in the Analyze chapter for
details about coloring etc.
User Guide to SIMCA
532

Figure 1. A typical PLS-Tree. Each branch of the tree corresponds to a PLS model
fitted to a sub-set of observations.
Calculation of PLS-Trees
Cross validation is used to terminate the branches of the tree, and to determine the
number of components of each cluster PLS model.
To accomplish a PLS-Tree the data matrices, X and Y, are first centered and scaled, as
usual. A PLS analysis is made of X and Y. Thereafter, the first X-score, t1, is used as
the dividing coordinate together with the Y-data (X and Y are sorted along t1). The
point on t1 is searched that divides X and Y in two parts, 1 and 2, such that the
following expression is minimized:
* (N
1
- N
2
)
2
/(N
1
+ N
2
)
2
+ (1 - ) * [ * (( V
Y1
+ V
Y2
)/ V
Y
) + (1 - ) * + ((V
t1
+
V
t2
)/V
t
)]
In the expression above, V denotes variance,
and , A and B in SIMCA, are two adjustable parameters.
The parameters A and B both run between 0 and 1. They regulate how the PLS models
are split (i.e., how the observations from one upper level PLS model are distributed
among two lower level models) according to the score t1, the Y-variable(s) or the
group size. The first parameter, A, sets the balance between the score t1 and the Y; the
closer to zero the more weight is attributed to the score t1. The second parameter, B,
takes into account the group size of the resulting clusters; the closer to zero the less
important it becomes to have equal group sizes in the dendrogram. In summary, this
means that a division along t1 is sought that minimizes the within groups variation and
hence maximizes the between group differences in t1 and Y.
The approach results in the row-wise splitting of data into a tree structure
(dendrogram) of PLS models, one split for each cluster (node in the dendrogram). The
dendrogram with the associated PLS models is called a PLS-Tree. When Y comprises
a discrete matrix with 1/0 columns corresponding to a number of predefined classes,
the result is a PLS classification tree.
Appendix A: Statistics
533
References cluster analysis
1. Kriegl, J.M., Eriksson, L., Arnhold, T., Beck, B., Johansson, E., and Fox, T.,
(2005), Multivariate Modeling of Cytochrome P450 3A4 Inhibition,
European Journal of Pharmaceutical Sciences, 24, 451-463.
2. Eriksson, L., Johansson, E., Mller, M., and Wold, S., (2000c), On the
Selection of Training Set in Environmental QSAR When Compounds are
Clustered, Journal of Chemometrics, 14, 599-616.
3. Breiman, L., Friedman, J.H., Olshen, R.A., and Stone, C.J., (1984)
Classification and Regression Trees, Wadsworth & Brooks / Cole Advanced
Books & Software, Monterey, CA.
4. Everitt, B.S., Landau, S., Leese, M., (2001), Cluster Analysis, Fourth Edition
Arnold Publishers, London.
5. Ward, J.H., (1963), Hierarchical grouping to optimize an objective function.
J. Am. Stat. Assoc., 58, 236-244.
Vectors available in SIMCA
Modeling using PCA, PLS, OPLS, and O2PLS results in a number of vectors
describing the model and its properties, and the residuals.
In the subsections that follow all vectors available in SIMCA are described. The
vectors are divided in groups according to their data type in the Plot/List dialogs.
- For vectors such as variables, scores, and distance to model, see the
Variables and Scores vectors subsection.
- For vectors such as observations, loadings, and Q2VX, see the Observations
and Loadings vectors subsection.
- For vectors pertaining to a component such as Q2, R2X, or SDt see the
Function of component vectors subsection.
- For lag vectors for models with lags defined in the workset, see the Function
of lags vectors subsection.
- For aligned vectors for batch evolution models, see the Aligned Vectors
subsection.
- For Out Of Control vectors for models in batch evolution models, see the
Batch Vectors - Out Of Control Summary subsection.
Some of the vectors are described with formulas in the Formulas and descriptions
section later in this chapter.
Note: All vectors available for PLS are also available (with the same
description) for OPLS and O2PLS unless otherwise stated. For OPLS and
O2PLS these common vectors generally refer to the predictive side.
OPLS and O2PLS specific vectors are described in the Orthogonal PLS modeling
subsection earlier in this chapter.
User Guide to SIMCA
534
Variables and Scores vectors
In the Plot/List tab dialogs, the vectors found in the Items box with data type
Variables and Scores are column vectors with the same shape as variables and scores,
i.e., with one element per observation.
See the table for the vectors available and their description in alphabetical order. The
rightmost column displays the plot/list/spreadsheet where the vector is available, when
applicable, in bold text if the vector is displayed by default.
Vector Description Displayed
CVgroups The cross-validation group that each
observation is assigned to.

Date/Time Variable in the dataset specified as
Date/Time Variable in the SIMCA import.
For more see the Formatting variable as
Date/Time subsection in the SIMCA
import chapter.
All line plots
displaying
observations,
for instance:
Home |
Scores | Line
Date/Time-PS Predictionset variable specified as
Date/Time Variable in the SIMCA import.
For more see the Formatting variable as
Date/Time subsection in the SIMCA
import chapter.
All line plots
displaying
observations,
for instance:
Predict |
Score PS|
Line
DModX

Distance to the model in X space (row
residual SD), after A components (the
selected model dimension), for the
observations used to fit the model. If you
select component 0, it is the standard
deviation of the observations with scaling
and centering as specified in the workset,
i.e., it is the distance to the origin of the
scaled coordinate system.
(A) = Absolute distance.
(N) = Normalized distance.
(M) = Mpow weighted residuals.
Home |
DModX
DModXPS Distance to the model in the X space (row
residual SD), after A components (the
selected dimension), for new observations
in the predictionset. Displaying component
0, it is the standard deviation of the
observations with scaling as specified in
the workset times v (correction factor for
workset observations, see the Absolute
distance to the model of an observation
in the workset subsection in Appendix
A: Statistics), i.e., it is the distance to the
origin of the scaled coordinate system.
Predict |
DModX PS
DModXPS+

Combination of DModXPS and tPS, when
the latter is declared different from the
workset observations.
Predict |
DModX PS
Appendix A: Statistics
535
Vector Description Displayed
DModY Distance to the model in the Y space (row
residual SD) after A components (the
selected model dimension) for the
observations used to fit the model. If you
select component 0, it is the standard
deviation of the observations with scaling
and centering as specified in the workset.
Line or Column
under the
DModY header
in Home |
DModX
DModYPS Distance to the model in the Y space (row
residual SD) after A components (the
selected model dimension) for
observations in the predictionset. If you
select component 0, it is the standard
deviation of the observations with scaling
and centering as specified in the workset.
Line or Column
under the
DModY header
in Predict |
DModXPS
Num Index number: 1, 2, 3 etc. All lists
displaying
observations,
for instance
Predict | Y PS
| List
ObsID Numerical observation identifiers, primary,
secondary, batch, or phase.

OLevX The leverage is a measure of the influence
of a point (observation) on the PC model
or the PLS model in the X space.
The observations leverages are computed
as the diagonal elements of the matrix H
0

after A dimensions.
H
0
= T[T'T]
-1
T'.

OLevY The leverage is a measure of the influence
of a point (observation) on the PLS model
in the Y space.
The observations leverages are computed
as the diagonal elements of the matrix H
y

after A dimensions.
H
y
= U[U'U]
-1
U'.

ORisk The observation risk is a sensitivity
measure and indicates the "risk" (the
influence) of including an observation in
the workset. ORisk is based on the Y
residual (in a PLS/OPLS/O2PLS model)
for a selected observation. It is computed
from the difference between the residual
standard deviation of the selected Y, when
the observation is and is not in the model.

User Guide to SIMCA
536
Vector Description Displayed
ORisk(pooled) The pooled ORisk is analogous to ORisk
but is valid across all Y variables in a
PLS/OPLS/O2PLS model rather than an
individual Y variable. It is computed from
the difference between the pooled residual
standard deviation of the Y variables,
when the observation is and is not in the
model.

PModX Probability of belonging to the model in the
X space, for observations used to fit the
model. Component 0 corresponds to a
point model, i.e., the center of the
coordinate system.
Observations with probability of belonging
of less than 5% are considered to be non-
members, i.e., they are different from the
normal observations used to build the
model.

PModY Probability of belonging to the model in the
Y space, for observations used to fit the
model. Component 0 corresponds to a
point model, i.e., the center of the
coordinate system.
Observations with probability of belonging
of less than 5% are considered to be non-
members, i.e., they are different from the
normal observations used to build the
model.

PModXPS Probability of belonging to the model in the
X space, for new observations in the
predictionset. Component 0 corresponds
to a point model, i.e., the center of the
coordinate system.
Observations with probability of belonging
of less than 5% are considered to be non-
members, i.e., they are different from the
normal observations used to build the
model.
Predict |
Prediction
List
PModXPS+ The same as PModXPS but based on
DModXPS+ instead of DModXPS.
Predict |
Prediction
List
Predict |
Classification
List
Appendix A: Statistics
537
Vector Description Displayed
PModYPS Probability of belonging to the model in the
Y space, for new observations in the
predictionset. Component 0 corresponds
to a point model, i.e., the center of the
coordinate system.
Observations with probability of belonging
of less than 5% are considered to be non-
members, i.e., they are different from the
normal observations used to build the
model.

SerrL Lower limit of the standard error of the
predicted response Y for an observation in
the workset.

SerrLPS Lower limit of the standard error of the
predicted response Y for a new
observation in the predictionset.
SerrLPS is always in original units, i.e.,
back transformed when Y was
transformed, when displayed in the
Prediction List.
Predict |
Prediction
List
SerrU Upper limit of the standard error of the
predicted response Y for an observation in
the workset.

SerrUPS Upper limit of the standard error of the
predicted response Y for a new
observation in the predictionset.
SerrUPS is always in original units, i.e.,
back transformed when Y was
transformed, when displayed in the
Prediction List.
Predict |
Prediction
List
t

Scores t, one vector for each model
dimension, are new variables computed as
linear combinations of X. They provide a
summary of X that best approximates the
variation of X only (PC model), and both
approximate X and predict Y
(PLS/OPLS/O2PLS model).
Home |
Scores

T2Range Hotelling's T2 for the selected range of
components. It is a distance measure of
how far away an observation is from the
center of a model hyperplane.
For OPLS/O2PLS the range is locked to
using first predictive to last Orthogonal in
X.
Home |
Hotelling's T2
T2RangePS Predicted Hotelling's T2 for the selected
range of components.
For OPLS/O2PLS the range is locked to
using first predictive to last Orthogonal in
X.
Predict |
Hotelling's
T2PS
User Guide to SIMCA
538
Vector Description Displayed
tcv X score t for the selected model
dimension, computed from the selected
cross validation round.
For PLS/OPLS/O2PLS tcv contains one
value per observation and for PCA one
value per observation and cross validation
round.
Analyze | CV
Scores
tcvSE Jack-knife standard error of the X score t
computed from all rounds of cross
validation.

to Orthogonal X score to of the X-part of the
OPLS/O2PLS model, for the selected
component. It summarizes the unique X
variation, i.e., the X variation orthogonal to
Y.
Home |
Scores
Home |
Scores | Orth
X
tocv Orthogonal X score to from the X-part of
the OPLS/O2PLS model, for a selected
model dimension, computed from the
selected cross validation round.
toPS Predicted orthogonal X score to of the X-
part of the OPLS/O2PLS model, for the
observations in the predictionset
Predict |
Scores PS
toPScv Matrix of cross validated predicted scores
ToPS for the predictionset. Available for
OPLS and O2PLS models.
tPS Predicted X score t, for the selected model
dimension, for the observations in the
predictionset.
Predict |
Score PS
tPScv Predicted X score t, for the selected model
dimension, computed from the selected
cross validation round.

tPScvSE Jack-knife standard error of the X score t,
for the observations in the predictionset,
computed from all rounds of cross
validation.

u

Scores u, one vector for each model
dimension, are new variables summarizing
Y so as to maximize the correlation with
the X scores t.
Analyze |
Inner Relation
ucv Y score u for the selected model
dimension, computed from the selected
cross validation round.

uo Orthogonal Y score uo of the Y-part of the
OPLS/O2PLS model, for the selected
component. It summarizes the unique Y
variation, i.e., the Y variation orthogonal to
X.
Home |
Scores | Orth
Y
Appendix A: Statistics
539
Vector Description Displayed
uocv Orthogonal Y score uo of the Y-part of the
OPLS/O2PLS model, computed from the
selected cross validation round.

VarDS Variable from the selected dataset, in
original units. Available after selecting a
DS in the Data box.

VarID Variable Identifier.
VarPS

X variable from the predictionset. Can be
displayed in transformed or scaled units.

XVar X variable from the workset. Can be
displayed in transformed or scaled units.

XVarPred A reconstructed variable from the workset.
For PLS and PCA models, an X variable
from the workset is reconstructed as
X=TP'. For OPLS models XVarPred
represents the X-values predicted from the
given Y-values.

XVarPredPS A reconstructed variable from the
predictionset. For PLS and PCA models,
an X variable from the predictionset is
reconstructed as X=TPS * P'. For OPLS
models XVarPredPS represents the X-
values predicted from the given Y-values.

XVarPS X variable from the predictionset. Can be
displayed in transformed or scaled units.
Predict |
Prediction
List
XVarRes X variable residuals for observations in the
workset, in original units. Can be displayed
in transformed or scaled units.

XVarResPS X variable residuals for observations in the
predictionset, in original units. Can be
displayed in transformed or scaled units.

XVarResPSSt X variable residuals for observations in the
predictionset, in standardized units
(divided by the residual standard
deviation). Can be displayed in
transformed or scaled units.

XVarResSt X variable residuals for observations in the
workset, in standardized units (divided by
the residual standard deviation). Can be
displayed in transformed or scaled units.

XVarResYRelated X variable residuals where the systematic
variation orthogonal to Y has been
removed. Available for OPLS and O2PLS.

User Guide to SIMCA
540
Vector Description Displayed
YPred Predicted values of Y variables for
observations in the workset, in original
units, i.e., back-transformed when
transformations are present. Can be
displayed in transformed or scaled units.
Home |
Observed vs.
Predicted
YPredcv Predicted values of the fitted Ys for
observations in the workset, computed
from the cross validation procedure.

YPredPS Predicted values for Y variables for
observations in the predictionset, in
original units, i.e. back transformed when
transformations are present. Can be
displayed in transformed or scaled units.
Predict | Y PS
Predict |
Prediction
List
YPredErrcv Prediction error of the fitted Ys for
observations in the workset, computed
from the cross validation procedure.

YPredPSConfInt-
YPredPSConfInt+
Lower/upper limit for the confidence
interval of predicted Ys from the
predictionset. The limit is calculated from
the cross validation and the confidence
level specified in model options.

YPredPScv Predicted values of the modeled Ys for
observations in the predictionset,
computed from the cross validation
procedure.

YPredPScvSE Jack-knife standard error of the prediction
of Y for observations in the predictionset,
computed from all rounds of cross
validation.

YVar Y variable from the workset. Can be
displayed in transformed or scaled units.
Home |
Observed vs.
Predicted
YVarPS Y variable from the predictionset. Can be
displayed in transformed or scaled units.
Predict | Y PS
YVarRes Y variable residuals for observations in the
workset, in original units. Can be displayed
in transformed or scaled units.

YVarResSt

Y variable residuals for observations in the
workset, in standardized units (divided by
the residual standard deviation). Can be
displayed in transformed or scaled units.

YVarResPS Y variable residuals for observations in the
predictionset, in original units. Can be
displayed in transformed or scaled units.

Appendix A: Statistics
541
Vector Description Displayed
YVarResPSSt Y variable residuals for observations in the
predictionset, in standardized units
(divided by the residual standard
deviation). Can be displayed in
transformed or scaled units.

Observations and Loadings vectors
In the Plot/List tab dialogs the vectors found in the Items box with data type
Observations and Loadings are row vectors with the same shape as an observation,
i.e., with one row per variable.
See the table for the vectors available and their description in alphabetical order. The
rightmost column displays the plot/list/spreadsheet where the vector is available, when
applicable, in bold text if the vector is displayed by default.
Vector Description Displayed
Batch VIP

The Batch Variable Importance plot (Batch
VIP) is available for batch level models and
displays the overall importance of the variable
on the final quality of the batch. With phases,
the plot displays the importance of a variable
by phase. With a PLS model, the Batch VIP
displays the plot for one y-variable at a time,
with a column per variable and per selected
phase.
Note: The Batch VIP is only available for
scores batch level datasets.
Batch |
Variable
Importance
c For every dimension in the PLS model there is
a c vector. It contains the Y loading weights
used to linearly combine the Y's to form the Y
score vector u. This means the c vector
actually expresses the correlation between the
Y's and the X score vector t.
Home |
Loadings
c(corr) Y loading weight c scaled as a correlation
coefficient between Y and u.
Home |
Loadings
Analyze |
Biplot
ccv Y loading weight c for a selected model
dimension, computed from the selected cross
validation round.

ccvSE Jack-knife standard error of the Y loading
weight c computed from the rounds of cross
validation.

User Guide to SIMCA
542
Vector Description Displayed
co Orthogonal Y loading weights co combine the
Y variables (first dimension) or the Y residuals
(subsequent dimensions) to form the scores
Uo.
These orthogonal Y loading weights are
selected so as to minimize the correlation
between Uo and T, thereby indirectly between
Uo and X. Available for OPLS and O2PLS
models.

cocv Orthogonal Y loading weights co from the Y-
part of the model, for a selected model
dimension, computed from the selected cross
validation round. Available for OPLS and
O2PLS models.

Coeff PLS/OPLS/O2PLS regression coefficients
corresponding to the unscaled and uncentered
X and Y. This vector is cumulative over all
components up to the selected one.
Home |
Coefficients

CoeffC PLS/OPLS/O2PLS regression coefficients
corresponding to the unscaled but centered X
and unscaled Y. This vector is cumulative over
all components up to the selected one.
Home |
Coefficients
CoeffCS

PLS/OPLS/O2PLS regression coefficients
corresponding to centered and scaled X, and
scaled (but uncentered) Y. This vector is
cumulative over all components up to the
selected one.
Home |
Coefficients
CoeffCScv PLS/OPLS/O2PLS regression coefficients
corresponding to the centered and scaled X
and the scaled (but uncentered) Y computed
from selected the cross validation round.

CoeffCScvSE Jack-knife standard error of the coefficients
CoeffCS computed from all rounds of cross
validation.

CoeffMLR PLS/OPLS/O2PLS regression coefficients
corresponding to the scaled and centered X
but unscaled and uncentered Y. This vector is
cumulative over all components up to the
selected one.
Home |
Coefficients
CoeffRot Rotated PLS/OPLS/O2PLS regression
coefficients corresponding to the unscaled and
uncentered X and Y. This vector is cumulative
over all components up to the selected one.
Home |
Coefficients
MPowX The modeling power of variable X is the
fraction of its standard deviation explained by
the model after the specified component.

Num Index number: 1, 2, 3 etc.
Appendix A: Statistics
543
Vector Description Displayed
ObsDS Observation in the dataset, selected in the
Data box, in original units.

ObsPS Observation in the current predictionset, in
original units. There is only one current
predictionset at a time although many can be
specified.

p Loadings of the X-part of the model.
With a PCA model, the loadings are the
coefficients with which the X variables are
combined to form the X scores, t.
The loading, p, for a selected PCA dimension,
represent the importance of the X variables in
that dimension.
With a PLS model, p expresses the importance
of the variables in approximating X in the
selected component.
Home |
Loadings
p(corr)

X loading p scaled as a correlation coefficient
between X and t.
Home |
Loadings
Analyze |
Biplot
pc X loading p and Y loading weight c combined
to one vector.
Home |
Loadings
pc(corr) X loading p and Y loading weight c scaled as
correlation coefficients between X and t (p) and
Y and u (c), and combined to one vector.
Home |
Loadings
Analyze |
Biplot
pccvSE Jack-knife standard error of the combined X
loading p and Y loading weight c computed
from all rounds of cross validation.

pcv X loading p for a selected model dimension,
computed from the selected cross validation
round.

pcvSE Jack-knife standard error of the X loading p
computed from all rounds of cross validation.

po Orthogonal loading po of the X-part of the
OPLS/O2PLS model. po expresses the unique
variability in X not found in Y, i.e, X variation
orthogonal to Y, in the selected component.
Home |
Loadings
Home |
Loadings |
Orth X
po(corr) Orthogonal loading po of the X-part of the
OPLS/O2PLS model, scaled as the correlation
coefficient between X and to, in the selected
component.
User Guide to SIMCA
544
Vector Description Displayed
pocv Orthogonal loading po of the X-part of the
OPLS/O2PLS model, for a selected model
dimension, computed from the selected cross
validation round.
poso Orthogonal loading po of the X-part and the
projection of to onto Y, so, combined to one
vector. Available for OPLS and O2PLS.
Home |
Loadings
Home |
Loadings |
Orth X
pq X loading weight p and Y loading weight q
combined to one vector. Available for OPLS
and O2PLS.
Home |
Loadings
Home |
Loadings |
Pred X-Y
q Loadings of the Y-part of the OPLS/O2PLS
model.
q expresses the importance of the variables in
approximating Y variation correlated to X, in
the selected component. Y variables with large
q (positive or negative) are highly correlated
with t (and X).
Home |
Loadings
Home |
Loadings |
Pred X-Y
qcv Y loading q for a selected model dimension,
computed from the selected cross validation
round. Available for OPLS and O2PLS models.
Q2VX, Q2VY

Predicted fraction, according to cross
validation, of the variation of the X (PCA) and Y
variables (PLS/OPLS/O2PLS), for the selected
component.
Home |
Component
Contribution
Q2VXcum,
Q2VYcum
Cumulative predicted fraction, according to
cross validation, of the variation of the X
variables (PCA model) or the Y variables
(PLS/OPLS/O2PLS model).
Home |
Summary of
Fit
qo Orthogonal loading qo of the Y-part of the
OPLS/O2PLS model.
qo expresses the unique variability in Y not
found in X, i.e, Y variation orthogonal to X, in
the selected component.
Home |
Loadings |
Orth Y
qocv Orthogonal loading qo of the Y-part of the
OPLS/O2PLS model, for a selected model
dimension, computed from the selected cross
validation round.
qor qo and r combined to one vector. Available for
OPLS and O2PLS.
Home |
Loadings |
Orth Y
Appendix A: Statistics
545
Vector Description Displayed
r R is the projection of uo onto X.
R contains non-zero entries when the score
matrix Uo is not completely orthogonal to X.
The norm of this matrix is usually very small
but is used to enhance the predictions of X.
Available for OPLS and O2PLS.
Home |
Loadings |
Orth Y
R2VX Explained fraction of the variation of the X
variables, for the selected component.
Home |
Component
Contribution
R2VXAdj Explained fraction of the variation of the X
variables, adjusted for degrees of freedom, for
the selected component.
Home |
Component
Contribution
R2VXAdjcum Cumulative explained fraction of the variation
of the X variables, adjusted for degrees of
freedom.
Home | X/Y
Overview
R2VXcum Cumulative explained fraction of the variation
of the X variables.
Home | X/Y
Overview
R2VY Explained fraction of the variation of the Y
variables, for the selected component.
Home |
Component
Contribution
R2VYAdj

Explained fraction of the variation of the Y
variables, adjusted for degrees of freedom, for
the selected component.
Home |
Component
Contribution
R2VYAdjcum

Cumulative explained fraction of the variation
of the Y variables, adjusted for degrees of
freedom.
Home | X/Y
Overview
R2VYcum Cumulative explained fraction of the variation
of the Y variables.
Home | X/Y
Overview
RMSEcv Root Mean Square Error, computed from the
selected cross validation round.
Analyze |
RMSECV
RMSEE Root Mean Square Error of the Estimation (the
fit) for observations in the workset.

RMSEP Root Mean Square Error of the Prediction for
observations in the predictionset.
Predict | Y PS
| Scatter
Predict | Y PS
| Line
S2VX Residual variance of the X variables, after the
selected component, scaled as specified in the
workset.

S2VY Residual variance of the Y variables, after the
selected component, scaled as specified in the
workset.

User Guide to SIMCA
546
Vector Description Displayed
so So is the projection of to onto Y.
So contains non-zero entries when the score
matrix To is not completely orthogonal to Y.
The norm of this matrix is usually very small
but is used to enhance the predictions of Y.
Available for OPLS and O2PLS models.
Home |
Loadings |
Orth X
VarID Numerical variable identifiers, primary or
secondary.
All lists
displaying
variables, for
instance Home
| Coefficients |
List
VIP Variable Influence on the Projection. It provides
the influence of every term in the matrix X on
all the Y's. Terms with VIP>1 have an above
average influence on Y. This vector is
cumulative over all components up to the
selected one.
Home | VIP
VIPcv VIP computed from the selected cross
validation round.

VIPcvSE Jack-knife standard error of the VIP computed
from all rounds of cross validation.

w

X loading weight that combine the X variables
(first dimension) or the X residuals (subsequent
dimensions) to form the scores t. This loading
weight is selected so as to maximize the
correlation between t and u, thereby indirectly
between t and Y.
X variables with large w's (positive or negative)
are highly correlated with u (and Y).
Home |
Loadings
w*

X loading weight that combines the original X
variables (not their residuals in contrast to w) to
form the scores t.
In the first dimension w* is equal to w.
w* is related to the correlation between the X
variables and the Y scores u.
W* = W(P'W)
-1

X variables with large w* (positive or negative)
are highly correlated with u (and Y).
Home |
Loadings
w*c X loading weight w* and Y loading weight c
combined to one vector.
Home |
Loadings
w*ccvSE

Jack-knife standard error of the combined X
loading weight w* and Y loading weight c
computed from all rounds of cross validation.

w*cv X loading weight w*, for a selected model
dimension, computed from the selected cross
validation round.

Appendix A: Statistics
547
Vector Description Displayed
w*cvSE Jack-knife standard error of the X loading
weight w* computed from all rounds of cross
validation.

wcv X loading weight w, for a selected model
dimension, computed from the selected cross
validation round.

wcvSE Jack-knife standard error of the X loading
weight w computed from all rounds of cross
validation.

wo Orthogonal loading weight wo of the X-part of
the OPLS/O2PLS model. It combines the X
residuals to form the orthogonal X score to.
This loading weight is selected so as to
minimize the correlation between to and u,
thereby indirectly between to and Y.

wocv Orthogonal loading weight wo of the X-part of
the OPLS/O2PLS model, for a selected model
dimension, computed from the selected cross
validation round.

Xavg Averages of X variables, in original units. If the
variable is transformed, the average is in the
transformed metric.

XObs X variables for the selected observation in the
workset in original units. Can be displayed in
transformed or scaled units.

XObsPred Reconstructed observations as X=TP from the
workset. Can be displayed in transformed or
scaled units.

XObsPredPS Reconstructed observations as X=TP from the
predictionset. Can be displayed in transformed
or scaled units.

XObsRes Residuals of observations (X space) in the
workset, in original units. Can be displayed in
transformed or scaled units.

XObsResPS Residuals of observations (X space) in the
predictionset, in original units. Can be
displayed in transformed or scaled units.

Xws Scaling weights of the X variables.
User Guide to SIMCA
548
Vector Description Displayed
Y-Related
Profiles
Displays the estimated pure profiles of the
underlying constituents in X under the
assumption of additive Y-variables.
Estimation includes a linear transformation of
the Coefficient matrix, Bp(Bp
T
Bp)
-1
, where Bp
is the Coefficient matrix using only the
predictive components to compute the
Coefficient matrix (i.e., the components
orthogonal to Y are not included in the
computation of Bp). Available for OPLS and
O2PLS models.
Analyze | Y-
Related
Profiles
Yavg Averages of Y variables, in original units. If the
variable is transformed, the average is in the
transformed metric.

YObs Y variables for the selected observation in the
workset in original units. Can be displayed in
transformed or scaled units.

YObsRes Residuals of observations (Y space) in the
workset, in original units. Can be displayed in
transformed or scaled units.

YObsResPS Residuals of observations (Y space) in the
predictionset, in original units. Can be
displayed in transformed or scaled units.

Yws Scaling weights of the Y variables.
Function of component vectors
For the Plot/List tab dialogs the vectors found in the Items box with data type
F(Component) are available per component. Some statistics apply to the whole
matrix; others are for selected variable, for each component.
See the table for the vectors available and their description in alphabetical order. The
rightmost column displays the plot/list/spreadsheet where the vector is available, when
applicable, in bold text if the vector is displayed by default.
Vector Description Displayed
Eig Eigenvalues of the X matrix. View | Model
Window
Iter Number of iterations of the
algorithm till convergence.
View | Model
Window
Num Index number: 1, 2, 3 etc.

Q2 Fraction of the total variation of
the X block (PCA) or the Y block
(PLS) that can be predicted by
each component.
View | Model
Window
Appendix A: Statistics
549
Vector Description Displayed
Q2(cum)progression Cumulative Q2 for the extracted
components, showing the
progression of cumulative values
for each added orthogonal
component in the OPLS/O2PLS
model, e.g. 1+0, 1+1, 1+2.
Available for OPLS/O2PLS
models with 1+x+0 components.
Home | Summary of
Fit
Q2cum Cumulative Q2 for the extracted
components.
View | Model
Window
Home | Summary of
Fit
Q2VX, Q2VY Predicted fraction, according to
cross validation, of the variation
of the X (PCA) and Y variables
(PLS/OPLS/O2PLS), for the
selected component.
Home | Summary of
Fit | Component
Contribution
Q2VXcum,
Q2VYcum
Cumulative predicted fraction,
according to cross validation, of
the variation of the X variables
(PCA model) or the Y variables
(PLS/OPLS/O2PLS model).
Home | Summary of
Fit | X/Y Overview
R2(cum)progression Cumulative fraction of Y
variation, showing the
progression of cumulative values
for each added orthogonal
component in the OPLS/O2PLS
model, e.g. 1+0, 1+1, 1+2.
Available for OPLS/O2PLS
models with 1+x+0 components.
Home | Summary of
Fit
R2VX Explained fraction of the variation
of the X variables, for the
selected component.
Home | Summary of
Fit | Component
Contribution
R2VXAdjcum Cumulative explained fraction of
the variation of the X variables,
adjusted for degrees of freedom.
Home | Summary of
Fit | X/Y Overview
R2VXcum Cumulative explained fraction of
the variation of the X variables.
Home | Summary of
Fit | X/Y Overview
R2VY Explained fraction of the variation
of the Y variables, for the
selected component.
Home | Summary of
Fit | Component
Contribution
R2VYAdjcum Cumulative explained fraction of
the variation of the Y variables,
adjusted for degrees of freedom.
Home | Summary of
Fit | X/Y Overview
R2VYcum Cumulative explained fraction of
the variation of the Y variables.
Home | Summary of
Fit | X/Y Overview
User Guide to SIMCA
550
Vector Description Displayed
R2X Fraction of the total variation of
the X block that can be explained
by each component.
View | Model
Window.
Home | Summary of
Fit
R2XAdj Explained fraction of the variation
of the X block, adjusted for
degrees of freedom, for the
selected component.
Home | Summary of
Fit
R2XAdjcum Cumulative explained fraction of
the variation of the X block,
adjusted for degrees of freedom.

R2Xcum Cumulative explained fraction of
the variation of the X block.
View | Model
Window.
Home | Summary of
Fit
R2Y Fraction of the total variation of
the Y block that can be explained
by each component.
View | Model
Window.
Home | Summary of
Fit
R2YAdj Explained fraction of the variation
of the Y block, adjusted for
degrees of freedom, for the
selected component.
Home | Summary of
Fit
R2YAdjcum Cumulative explained fraction of
the variation of the Y block,
adjusted for degrees of freedom.

R2Ycum Cumulative explained fraction of
the variation of the Y block.
View | Model
Window.
Home | Summary of
Fit
RMSEcv Root Mean Square Error,
computed from the selected
cross validation round.
Analyze | RMSECV
RMSEcv-
progression
Root Mean Square Error showing
the progression of RMSEcv
values for each added orthogonal
component in the OPLS/O2PLS
model, e.g. 1+0, 1+1, 1+2.
Available for OPLS/O2PLS
models with 1+x+0 components.
Analyze | RMSECV
S2X Variance of the X block. For
component number A, it is the
residual variance of X after
component A.

Appendix A: Statistics
551
Vector Description Displayed
S2Y Variance of the Y block. For
component number A, it is the
residual variance of Y after
component A.

SDt Standard deviation of the X
scores, T.

SDu Standard deviation of the Y
scores, U.

SSX Sum of squares of the X block.
For component number A, it is
the X residual Sum of Squares
after component A.

SSY Sum of squares of the Y block.
For component number A, it is
the Y residual Sum of Squares
after component A.

YPredErrcvSE Jack-knife standard error of the
prediction error of the fitted Ys for
observations in the workset,
computed from the cross
validation rounds.

Function of lags vectors
For the Plot/List tab dialogs vectors found in the Items box with data type F(Lags) are
available as functions of lags.
See the table for the vectors available and their description in alphabetical order.
Vector Description
CoeffCScvSELag Jack-knife standard error on the coefficients as a function of
lag, computed from all cross validation rounds.
CoeffCSLag Coefficients (for scaled and centered data) of a lagged variable
x, for a selected Y as a function of lags.
Num Index number: 1, 2, 3 etc.
pLag X loading p of a lagged variable X, as a function of lags.
wLag X loading weight w of a lagged variable as a function of lags
w*Lag X loading weight w* of a lagged variable as a function of lags
VIPLag VIP of a lagged variable X as a function of lags.
Aligned Vectors
In the Plot/List tab dialogs the vectors found in the Items box with data type Aligned
Vectors are available for vectors aligned to median length. The average or standard
deviation of the vector can be displayed, and selected batches.
User Guide to SIMCA
552
Aligned vectors are available for batch evolution models.
Note: All aligned vectors are named '(Aligned)' in addition to the vector
name.
See the table for the vectors available and their description in alphabetical order. The
rightmost column displays the plot/list/spreadsheet where the vector is available.
Vector Description Displayed
DModX DModX vector aligned to median
length. See DModX description
previously.
Batch | DModX BCC.
DModXPS DModXPS vector aligned to median
length. See DModXPS description
previously.
Batch | DModX PS
BCC.
Num Index vector: 1, 2, 3 etc.
t t vector aligned to median length. See t
description previously.
Batch | Scores BCC.
T2Range T2Range vector aligned to median
length. See T2Range description
previously.
Batch | Hotelling's T2
BCC.
T2RangePS T2RangePS vector aligned to median
length. See T2RangePS description
previously.
Batch | Hotelling's
T2PS BCC.
Time/Maturity Time or Maturity variable determining
the end point of a Batch/Phase and
used as Y in the observation level
models. This variable is used to align
Batch/Phase to the median length.
All aligned batch
control charts.
tPS tPS vector aligned to median length.
See tPS description previously.
Batch | Scores PS
BCC.
XVar XVar vector aligned to median length.
See XVar description previously.
Batch | Variable BCC.
XVarPS XVarPS vector aligned to median
length. See XVarPS description
previously.
Batch | Variable PS
BCC.
YPred YPred vector aligned to median length.
See YPred description previously.
Batch | Observed vs.
Time/Maturity BCC.
YPredPS YPredPS vector aligned to median
length. See YPredPS description
previously.
Batch | Observed vs.
Time/Maturity PS
BCC.
YVar YVar vector aligned to median length.
See YVar description previously.
Batch | Observed vs.
Time/Maturity BCC.
YVarPS YVarPS vector aligned to median
length. See YVarPS description
previously.
Batch | Observed vs.
Time/Maturity PS
BCC.
Appendix A: Statistics
553
Batch Vectors
In the Plot/List tab dialogs the vectors found in the Items box with data type Batch
Vectors are the Out of Control Summary vectors, OOCSum, for all vectors displayed
in the batch control charts. That is: DModX, DModXPS, T2Range, T2RangePS, t, tPS,
XVar, XVarPS, YPred, and YPredPS. The OOCSum vectors are all named as the
mother vector with the suffix AlignedOOCSum.
The OOCSum for a given vector is the ratio of the area of the vector outside the
control limits to the total area inside the limits, for each batch. The OOCSum is then
expressed as a percent of the area inside the limits. With no part of the batch outside
the limits, OOCSum=0.
OOCSum is always computed on aligned vectors, both for workset and predictionset
vectors.
The OOCSum vector is default displayed using the default limits.
Change the limits as desired under High limit and Low limit after adding the series.


Vector Description
DModXAlignedOOCSum For a plot of DModX vs. Maturity for a batch in the
workset the out of control (OOC) sum is the area
outside the control limits expressed as a
percentage of the total area. The summation is
made relative to a DModX vector aligned to
median length.
DModXPSAlignedOOCSum For a plot of DModX vs. Maturity for a batch in the
predictionset the out of control (OOC)
sum is the area outside the control limits
expressed as a percentage of the total area.
T2RangeAlignedOOCSum For a plot of T2Range vs. Maturity for a batch in
the workset the out of control (OOC) sum is the
area outside the control limits expressed as a
percentage of the total area. The summation is
made relative to a T2Range vector aligned to
median length.
User Guide to SIMCA
554
Vector Description
T2RangePSAlignedOOCSum For a plot of T2Range vs. Maturity for a batch in
the predictionset the out of control (OOC) sum is
the area outside the control limits expressed as a
percentage of the total area. The summation is
made relative to a T2RangePS vector aligned to
median length.
TAlignedOOCSum For a plot of scores (t) vs. Maturity for a batch in
the workset the out of control (OOC) sum is the
area outside the control limits expressed as a
percentage of the total area. The summation is
made relative to a t vector aligned to median
length.
TPSAlignedOOCSum For a plot of scores (t) vs. Maturity for a batch in
the predictionset the out of control (OOC) sum is
the area outside the control limits expressed as a
percentage of the total area. The summation is
made relative to a tPS vector aligned to median
length.
XVarAlignedOOCSum For a plot of an X-variable vs. Maturity for a batch
in the workset the out of control (OOC) sum is the
area outside the control limits expressed as a
percentage of the total area. The summation is
made relative to an XVar vector aligned to median
length.
XVarPSAlignedOOCSum For a plot of an X-variable vs. Maturity for a batch
in the predictionset the out of control (OOC) sum is
the area outside the control limits expressed as a
percentage of the total area. The summation is
made relative to an XVarPS vector aligned to
median length.
YPredAlignedOOCSum For a plot of YPred vs. Maturity for a batch in the
workset the out of control (OOC) sum is the area
outside the control limits expressed as a
percentage of the total area. The summation is
made relative to an YPred vector aligned to
median length.
YPredPSAlignedOOCSum For a plot of YPred vs. Maturity for a batch in the
predictionset the out of control (OOC) sum is the
area outside the control limits expressed as a
percentage of the total area. The summation is
made relative to an YPredPS vector aligned to
median length.

Appendix A: Statistics
555
Formulas and descriptions
Fitting a PCA, PLS, OPLS, or an O2PLS model generates all the vectors listed in the
Vectors available in SIMCA section earlier in this chapter. This section describes and
provides formulas for a selection of those vectors. The descriptions for PLS vectors are
also true for the vectors created when fitting OPLS/O2PLS models unless otherwise
stated. When fitting with OPLS/O2PLS a few additional vectors are calculated and
these are listed in the same tables.
All vectors that have a 'Y' in the name refer to PLS/OPLS/O2PLS models.
SIMCA will iteratively compute one component at a time, that is: one vector each of
X-scores t, Y-scores u (for PLS/OPLS/O2PLS), weights w and c (for
PLS/OPLS/O2PLS), and loadings p.
The components are calculated in descending order of importance.
Prediction vectors
After fitting a model, this model can be used for predictive purposes for a given set of
observations.
The following vectors are described in this section:
- R2Y, R2X, R2Yadj, R2Xadj
- R2V and R2Vadj
- Q2 and Q2V
- Q2(cum) and Q2V(cum)
- Leverages
- RSD of observations and variables
- Modeling power - MPowX
- Hotelling's T2
- Missing values correction factor
- Score and loading vectors
- Distance to the model
- Variable importance, VIP
- Standard Error on the predicted values
- Coefficients
R2Y, R2X, R2Yadj, R2Xadj
R
2
Y and R
2
X display the fraction of the sum of squares for the selected component, SS
explained.
R
2
Yadj, R
2
Xadj display the variance explained in the model, that is, SS explained
corrected for degrees of freedom
User Guide to SIMCA
556
R2V and R2Vadj
For every variable in the model, the fraction of SS (R
2
V) or variance (R
2
Vadj)
explained can be displayed. This is computed for both the current component and
accumulated over all PC or PLS components. For response variables Y, this
corresponds to R
2
(the multiple correlation coefficient), the goodness of fit.
Q2 and Q2V
Q
2
: The fraction of the total variation of X or Y that can be predicted by a component,
as estimated by cross-validation.
Q
2
is computed as:
Q
2
= (1.0 - PRESS/SS)
Q
2
V: The fraction of the variation of a variable, e.g. x
k
or y
m
, that can be predicted by
a component, as estimated by cross-validation.
Q
2
V is computed as:
Q
2
VX = (1.0 - PRESS/SS)
k

Q
2
VY = (1.0 - PRESS/SS)
m

where Q
2
VX is available for both PCA and PLS, but Q
2
VY only for PLS.
When using SIMCA with cross validation, Q
2
and the limit are displayed in the model
window and the significance column informs if the component is significant or not and
according to which rule.
For details about the cross validation rules, see the Cross validation section later in
this chapter.
Q2(cum) and Q2V(cum)
The cumulative Q
2
for the extracted components is computed as:
Q
2
(cum) = (1.0 - (PRESS/SS)
a
)
where
[a = 1, ...A]
(PRESS/SS)
a
= the product of PRESS/SS for each individual component a.
When a component is insignificant the value PRESS/SS is truncated for that component
to 0.1, when that value is smaller than 0.1.
The cumulative Q
2
V of a variable is similarly computed as:
Q
2
V(cum) = (1.0 - (PRESS/SS)
ka
)
where
[a = 1, ...A]
When responses have large values of Q
2
V(cum) accumulated over all dimensions, the
model for this variable is good. An accumulated value larger than about 0.5 can be
considered large.
Appendix A: Statistics
557
Leverages
The leverage is a measure of the influence of a point (observation) on the PC or PLS
model. Leverage (OLev) is proportional to Hotelling's T
2
. Observations with high
leverages can exert a large influence on the model. A high leverage observation falling
near a component axis reinforces the model. A high leverage observation lying far
away from a component line causes a rotation of the model.
SIMCA computes the observations leverages in the X and Y spaces as the diagonal
elements of the matrices H
0
and H
y
, respectively:
H
0
= T(T'T)
-1
T'
H
Y
= U(U'U)
-1
U'
RSD of observations and variables
The residual standard deviation (RSD) can be computed for observations and variables.
The RSD of an observation in the X or Y space (rows in E and F) is proportional to the
observation distance to the hyper plane of the PLS model in the corresponding space
(DModX and DModY).
The RSD of an X variable (columns in E) relates to its relevance in the PC or PLS
model. The RSD of a Y variable (columns in F) is a measure of how well this response
is explained by the PLS model.
Relevance of variables
The importance of a variable in a PC model is indicated by the size of its explained
Sum of Squares (R
2
X) or variance (R
2
Xadj).
A variable is relevant if its modeling power = 1.0. Variables with low modeling power,
i.e., around (A/K) are of little relevance (A = number of model dimensions, K= number
of variables).
Modeling power - MPowX
The modeling power of a variable is defined as its explained standard deviation:
Mpow
k
= 1.0 - SV
k
/SV
0k

where
SV
k
= residual standard deviation (RSD) of variable x
k
.
SV
0k
= initial standard deviation of variable k. SV
0k
= 1.0 if the variable has been
autoscaled.
Hotelling's T2
T2Range is basically calculated as the sum over the selected range of components of
the scores in square divided by their standard deviations in square, provided the
variables were centered. Hence, T2Range is the distance in the model plane (score
space) from the origin, in the specified range of components. This means that the same
component is selected as 'From' and 'To' components, the plot displays the Hotelling's
T2 for that component.
For OPLS and O2PLS the range is locked to 'From' the first predictive 'To' the last
orthogonal in x.
User Guide to SIMCA
558
The Hotelling's T
2
for observation i, based on A components is:
T
i

2
= ((t
ia
- t
avg
)
2
/ s
ta
2
)
where
the summation is done over the range of the selected components,
s
2
ta
= Variance of t
a
according to the class model.
T
i
2
* (N - A) / A(N - 1)
is F distributed with A and N-A degrees of freedom.
N = Number of observations in the workset.
A = Number of components in the model or the selected number of components.
Hence if
T
i
2
> A(N - 1) / (N - A) * F
critical
(p=0.05)
then observation i is outside the 95% confidence region of the model.
Ellipse in score plot
The confidence region for a two dimensional score plot of dimension a and b is an
ellipse with axis:
\(s
2
ta or b
*F(2,N-2,o) * 2 * ((N -1)/(N-2)))
The significance level is by default 0.05. On the File tab, click Project Options and in
the Fit tab change the default level if desired.
To not display the ellipse, see the Limits and Regions and Limits subsections in the
Plot and list contextual tabs chapter.
Missing values correction factor
The NIPALS algorithm has been modified for data containing missing values. A limit
has been introduced on the correction factor used with missing values. This stabilizes
both the algorithm and the solution. A value of 3.0 for the limit on the correction factor
for both the scores and the loadings has generally been found to be appropriate.
Missing values and PCA
A very important problem in industrial applications of PCA, such as multivariate
statistical process control applications, is the estimation of scores when the observation
vector has missing data. SIMCAs approach to score calculations for such incomplete
observations are based on methods described in Nelson, P.R.C., Taylor, P.A.,
MacGregor, J.F., Missing Data Methods in PCA and PLS: Score Calculations with
Incomplete Observation, Chemometrics and Intelligent Laboratry Systems, 35, 45-65,
1996.
SIMCA uses different approaches to compute t and tPS for incomplete observations
and if the proportion of missing data is large there can be numerical differences
between t and tPS.
Appendix A: Statistics
559
Score and loading vectors
The score and loading plots available for plotting are described below.
Note that the score and loading plots complement each other. The position of an
observation in a given direction in a score plot is influenced by variables lying in the
same direction in the loading plot.
All of these score plots described here will reveal:
- Groups
- Trends
- Outliers
- Similarity
T scores (X): t1 vs. t2, ...
T score plots are windows in the X space displaying the observations as situated on the
projection plane or hyper plane.
With a 2-dimensonal score plot SIMCA draws the elliptical confidence interval based
on Hotelling's T
2
.
U scores (Y): u1 vs. u2, ...
U score plots are windows in the Y space, displaying the observations as situated on
the projection plane or hyper plane.
T vs. U scores (X&Y): t1 vs. u1, ...
T vs U score plots display the observations in the projected X(T) and Y(U) space, and
show how well the Y space correlates to the X space.
P loadings (X): p1 vs. p2,...
P loading plots show the importance of the X variables in the approximation of the X
matrix.
W loadings (X): w1 vs. w2, ...
W loading plots display the correlation between the X variables, in the first dimension,
or the residuals of the X variables in subsequent dimensions, and the Y or Y residuals
scores U(Y).
The w's are the weights that combine the X variables (first dimension) or the residuals
of the X variables (subsequent dimensions) to form the scores t. These weights are
selected so as to maximize the covariance between T and U, thereby indirectly T and
Y.
X variables with large w's (positive or negative) are highly correlated with U(Y).
Variables with large w's are situated far away from the origin (on the positive or
negative side) on the plot.
User Guide to SIMCA
560
W* loadings (X): w*1 vs. w*2, ...
For every PLS dimension the W
*
's are the weights that combine the original X
variables (not their residuals as with w) to form the scores t.
W
*
is computed as follows:
W* = W(P'W)
-1

From this formula we can see that w
*
is equal to w in the first dimension.
Also, when P is equal to W (for all dimensions), W
*
is equal to W because W'W (and
PW) is equal to 1.
Relation of W* to the PLS regression coefficients
For a given response Y, the PLS regression coefficients are (for the model Y = XB):
B = W(P'W)
-1
C'
or
B = W*C' and Y = XW*C'
Hence, W
*
is directly related to the PLS regression coefficients.
C loadings (Y): c1 vs. c2, ...
C loading plots display the correlation between the Y variables and T(X). The c's are
the weights that combine the Y variables with the scores u, so as to maximize their
correlation with X. Y variables with large c's are highly correlated with T(X).
WC or W*C loadings (X&Y): wc1 vs. wc2, ...
WC or W*C loading plots show both the X-weights (w or w
*
) and Y-weights (c), and
thereby the correlation structure between X and Y. One sees how the X and Y
variables combine in the projections, and how the X variables relate to Y.
TPS scores (X): tPS1 vs. tPS2, ...
The plots On the Predict tab classify (new) observations in the predictionset with
respect to a PC or PLS model and predicts responses (values of Y variables) for (new)
observations (PLS).
For a given set of observations, SIMCA computes the predicted scores: tPS. With
missing data in the new observation (denoted by z below), these scores are calculated
by means of PLS and the model below, according to Nelson, Taylor, and MacGregor
(see reference list).
z = P*tPS + e
Distance to the model
This section describes the different distances to the model available in SIMCA for
observations in the workset or in the predictionset.
- Absolute distance to the model of an observation in the workset
- Normalized Distance to the model of an observation in the workset
- Distance to the model of new observations in the predictionset
- MPow weighted distance to the model
Appendix A: Statistics
561
- Critical distance to the model
- Membership significance level
Absolute distance to the model of an observation in the
workset - DModX, DModY(Absolute)
For observations in the workset, SIMCA computes the observation distance to the
model in the X space, DModX, and in the Y space, DModY (for PLS models). These
distances are equivalent to X and Y residual standard deviations, respectively.
DModX absolute
DModX of an observation in the workset (i.e., that was part of the model) is computed
as:
s
i
= sqrt(e
ik
2
/ (K-A))* v
The summation is made over the X variables (k) and e
ik
are the X-residuals of
observation i.
v is a correction factor (function of the number of observations and the number of
components) and is slightly larger than one.
This correction factor takes into account the fact that the distance to the model
(DModX) is expected to be slightly smaller for an observation that is part of the
workset since it has influenced the model.
DModY absolute
DModY of an observation in the workset (i.e., that was part of the model) is computed
as:
s
i
= sqrt(f
im
2
/ M)
The summation is made over the Y variables (m) and f
im
are the Y residuals of
observation i.
DModY absolute for OPLS/O2PLS
DModY of an observation in the workset, when the fit method is OPLS/O2PLS is
computed as:
s
i
= sqrt(f
im
2
/ M) * v
The summation is made over the Y variables (m) and f
im
are the Y residuals of
observation i.
Normalized distance to the model of an observation in the
workset - DModX, DModY (Normalized)
The normalized distance to the model is the observation absolute DModX or DModY
divided by the pooled RSD of the model, s
0
, in the X space and in the Y space.
The pooled RSD for the X space is calculated as:
s
0
= \(e
ik
2
/ ((N-A-A
0
)*(K-A)))
A
0
= 1 if model is centered, 0 otherwise.
DModX normalized
DModX normalized = s
i
/s
0

where
s
i
is the absolute DModX.
User Guide to SIMCA
562
s
0
is the pooled RSD for the X space.
The pooled RSD for the Y space is calculated as:
s
0
= \(f
im
2
/ ((N-A-A
0
)*M))
DModY normalized
DModY normalized = s
i
/s
0

where
s
i
is the absolute DModY.
s
0
is the pooled RSD for the Y space.
Absolute distance to the model of new observations in the
predictionset - DModXPS, DModYPS (Absolute)
The absolute distance to the model DModXPS of a new observation is computed as the
observation RSD:
s
i
= \sqrt(e
ik
2
/ (K-A))
The difference in the formula in comparison with the calculation of DModX is that the
correction factor in not present.
The distance to the model in the Y space, DModYPS, of a new observation (i.e., that
was not part of the model) is computed in the same way as DModY, that is, for an
observation that was part of the model:
s
i
= \(f
im
2
/ M)
Normalized distance of a new observation to the model -
DModXPS, DModYPS (Normalized)
The normalized distance to the model is the observation absolute DModXPS or
DModYPS divided by the pooled RSD of the model (not the predictionset) s
0
:
Normalized DModXPS = s
i
/s
0

where
s
0
= \(e
ik
2
/ ((N-A-A
0
)*(K-A))
Normalized DModYPS = s
i
/s
0

where
s
0
= \(f
im
2
/ ((N-A-A
0
)*M)
A
0
= 1 if model is centered, 0 otherwise.
Distance to the model augmented - DModXPS+ (Absolute)
or (Normalized)
With predictions, the distance to the model can be augmented with a term measuring
how far outside (d) the acceptable model domain the projection of the observation
falls. The result is called DModXPS+. Hence, for an observation in the predictionset
(hence the PS label), this is:
DModXPS+ = [DModXPS
2
+ d
2
]
1/2

Where

d = d0
i
2
*S2X
i
/(SDt(i))
2
for components i = 1 to A.
Appendix A: Statistics
563
d0 is the distance between the score value t and the score limit for component
i for an observation in the predictionset. If t is within the limits, d0 is zero.
The score limits are:
Upper limit: Max(t) + SDt/2
Lower Limit: Min(t) - SDt/2
S2X
i
is the residual variance.
Here the distance from the projected point of the observation to the model domain is
re-expressed in the same units as the residuals by multiplication by the RSD of the
workset (s0) and division by the pooled standard deviation of the score vectors in
which the projected point is outside the acceptable interval.
The DModXPS+ can be expressed in absolute or normalized, and weighted by the
modeling power, just like DModXPS.
MPow weighted distance to the model - DModX, DModXPS,
DModXPS+ (Abs/Norm and weighted)
With MPow weighted DModX/DModXPS/DModXPS+, that is, distance to the model
for the X-block weighted by the modeling power, the observation RSD is computed
with weighted residuals, where the weights are the modeling powers of the variables.
Critical distance to the model
The critical limit for the distance to model measure (DCrit) is calculated by an inverse
cumulative F-distribution function. The three input parameters for this function are the
significance level and the degrees of freedom of the numerator and the denominator of
the F-distribution. In this case, the numerator is the degrees of freedom of the
observations in the training set (DF_obs) and the denominator is the degrees of
freedom of the model (DF_mod). DCrit is equal to the value F* in the figure below
where the blue area is the probability p that an observation will be outside the DCrit
limit. For example, if the significance level is set to 0.95, typically 95% of the
observations would have a DModX value below DCrit. The exact shape and scale of
the curve is determined by the degrees of freedom. The F* value can be extracted from
statistical tables of the F-distribution.

In SIMCA, the value of the degrees of freedom for the model is calculated as:
DF_mod = \((N-C-A)*(K-A))
where
N = number of observations
C = 1 if all X variables are centered, else 0
User Guide to SIMCA
564
A = number of components
K = number of X variables
(DF_mod cannot be smaller than 1)
The value of the degrees of freedom for the observations is calculated as:
If K > DF_mod
DF_obs = (M + Q A) / CORR
else
DF_obs = (M A) / CORR
where
M = the smallest value of [K, 100, DF_mod]
Q = square root of (K-DF_mod)
CORR = N / (N-C-A)
(N-C-A and DF_obs cannot be smaller than 1).
The expressions above result in a normalized DCrit value. If the critical limit is to be
presented in absolute values, DCrit is multiplied by the residual variance in X.
Membership significance level
As default, observations with a probability of class membership less than 5% are
considered non-members.
Variable importance, VIP
SIMCA computes the influence on Y of every term (x
k
) in the model, called VIP
(variable importance in the projection). VIP is the sum over all model dimensions of
the contributions VIN (variable influence). For a given PLS
dimension, a, (VIN)
ak
2
is equal to the squared PLS weight (w
ak
)
2
of that term,
multiplied by the explained SS of that PLS dimension.
The accumulated (over all PLS dimensions) value,
VIP
ak
2
= (VIN)
k
2

where the summation is made over a = 1 to A.
This value is then divided by the total explained SS by the PLS model and multiplied
by the number of terms in the model.
The final VIP is the square root of that number.
The formula can also be expressed as:
VIP
Ak
= \((w
ak
2
* (SSY
a-1
- SSY
k
))* (K/( SSY
a
- SSY
A
)))
The Sum of squares of all VIP's is equal to the number of terms in the model hence the
average VIP is equal to 1.
One can compare the VIP of one term to the others. Terms with large VIP, larger than
1, are the most relevant for explaining Y.
Appendix A: Statistics
565
Standard Error and Jack-knifing uncertainties
SIMCA computes standard errors on the predicted Y values in two different ways:
- According to Hskuldsson resulting in the Serr vectors.
- From Jack-knifing resulting in error bars in plots.
Standard error according to Hskuldsson - Serr
The variance of the prediction (Yhat) for a given response y at a point X
0
according to
Hskuldsson, is computed as:
V(Yhat) = (1/N + t
0
(T'T)
-1
t
0
' )
2

where
t
0
are the scores t of observation X
0