PEAXACT 5 Release Notes

Release Highlights • 5.0 • 5.1 • 5.2 • 5.3 • 5.4 • 5.5 • 5.6 • 5.7 • 5.8 • 5.9 • 5.10
Upgrade Notes
Compatibility Considerations

Consider checking your license before you upgrade.

5.10, 2024-10-30 Download Now

• IMPROVEDSupport for reading exact timestamps from Keit SPC-files.
• IMPROVEDSupport for reading exact timestamps from tec5 SPC-files saved by MultiSpec Pro II v1.5.1359.2024 or later.
• IMPROVED(AppServer) Error messages now contain a "caused by"-reason if available.
• FIXED3D surface plot not lit correctly in some cases when switching between features.
• FIXED(Analysis Report Windows) Incorrect sample file names in Data Cursors.
• FIXED(Calibration/Validation Report Window) In case of groupwise calibrations, switching between reports may not update the group name in plot titles.
• FIXED(License Activation) Downloading an invalid license still activates the license for the current host.
• FIXED(Hard Model) Poor fit in rare case when the signal (after pretreatments) is constant.
• FIXEDWindows may be opening / positioned outside the visible screen.
• FIXEDSeveral minor improvements and fixes.

5.9, 2023-12-10

• NEWEasy way to start the free PEAXACT Viewer on computers without a license.
• IMPROVEDLoaded model and sample files now get automatically sorted by filename. This does not apply to models/samples loaded from data table files or session files.
• IMPROVED(Data Inspector) On saving a filtered table, the user can choose to save the full table instead (prevention of data loss).
• IMPROVED(Data Inspector) On re-opening the Data Inspector, the user can choose to apply previous table changes first (prevention of data loss).
• IMPROVED(Data Inspector) Timestamp to Time convesion is more accurate, and a target time can be specified now.
• IMPROVED(Data Inspector) Non-contiguous cells can now be copied from the table, provided that the selection is rectangular.
• IMPROVED(Report Windows) Reopening a report window that is already open attempts to update the view with new results instead of resetting the view.
• IMPROVED(Calibration/Validation Report Window) Selection Tool now works in Differences Plot.
• IMPROVED(Validation Report Window) The RMS Residuals and Mahalanobis Distance plots now show limits (like plots during calibration).
• CHANGEDTransmittance > Absorbance transformation now allows TR values > 1 (100%).
• CHANGED(Data Inspector) When adding or renaming features, leading and trailing whitespaces are removed from the name.
• CHANGED(Main Window) When adding or renaming model components, leading and trailing whitespaces are removed from the name.
• CHANGED(Classification Setup) Options "centered" and "scaled" for Database Lookup have been replaced by new option "normal", which is more consistent in calculating class similarities. Old models using "centered" or "scaled" are still supported, but new models can only be created with option "normal".
• CHANGEDRemoved confirmation dialog before performing predictions with a groupwise calibration model. If available, groupwise prediction is now enabled by default.
• FIXEDWrong outlier statistics for groupwise predictions. [10]
• FIXED(Data Inspector) Editing a table with active column filters may freeze the application.
• FIXED(Calibration Window) Error when trying to export the table report to a file.
• FIXED(Classification Window) Error when trying to export the table report to a file.
• FIXED(Classification Window) Wrong colorbar ticks labels for numeric features.
• FIXED(Data Inspector) Trying to rename features to Usage or Quality (which is not possible) incorrectly changes names in the feature list.
• FIXED(Data Inspector) 3D surface plot may be lit incorrectly, e.g., when X2 = Timestamp.
• FIXEDIncorrect sorting of strings containing the number 0.
• FIXEDError when loading empty text files.
• FIXED(Validation Report Window) When using a groupwise calibration model, the RMS Residuals vs True Plot and Mahalanobis Distances vs True Plot may not display all training samples.
• FIXEDStarting PEAXACT with invalid command line arguments may crash the application.
• FIXED(Progress Window) When processing a lot of samples, the progress message may not update for every sample.
• FIXED(Create Representative Sample) Changing the number of considered samples has no effect.
• FIXEDExported plots use incorrect font size for data tips.
• FIXED(Data Inspector) Error in 3D Samples plot when there is no data within global range.
• FIXED(Data Inspector) Crash when trying to export pretreatments while refreshing plots.
• FIXEDColor bar labels may not update correctly when the color range is changed.
• FIXED(Data Inspector) Switching back and forth between Table Editor and Plotter disables any active tool.

5.8, 2023-02-03

• NEWAdded support for LightField .spe files.
• IMPROVEDLinear Hard Model fitting now uses an improved algorithm. Note that compared to the previous version, this may lead to slightly different numerical results in fitted component weights (in the order of 1e-6).
• IMPROVEDLoading sample files (Data > Load... or Drag&Drop) is faster.
• IMPROVEDReading from sample files is faster.
• IMPROVEDOpening the Data Inspector is faster.
• IMPROVEDSorting samples (in the Samples Panel) is faster.
• IMPROVEDThe Join Table Tool is significantly faster.
• IMPROVEDThe Selection Tool now supports single-click selection of objects.
• IMPROVEDThe Component Fit Preview is not turned off when Hard Model settings get changed.
• IMPROVEDThe Calibration/Classification Report Window does not reset the current choice (function or rank) when creating or editing alternative calibrations/classifications.
• IMPROVEDThe Classification Setup Dialog now warns if class names only differ by leading/trailing spaces or upper/lower case.
• IMPROVEDThe Save File Dialog for reports (plots/tables) now remembers the last file format.
• IMPROVEDThe Open/Save File/Folder Dialogs now honor directory junctions.
• IMPROVED(ProcessLink) The default data block read from OPUS files is now a spectrum (AB, TR, ...) instead of the first block found in the file.
• IMPROVED(AppServer) The AppServer .NET API is now compatible with .NET Core & .NET 5+.
• FIXEDRMS residuals of Hard Model Fits may get calculated too large. [8]
• FIXEDIncorrect NMR preprocessing. [9]
• FIXEDMCR-ALS error when samples have less than 30 data points.
• FIXED(Data Inspector) The correlation plot may not get displayed correctly, often related to having Alignment enabled in the Pretreatment Model.
• FIXED(Data Inspector) Error when adding table columns if samples have been relocated.
• FIXED(Data Inspector) Error when saving a table without features.
• FIXED(Data Inspector) The table filter dropdown menu may not contain all possible items.
• FIXED(Data Inspector) Wrong tooltip text for the Save tool.
• FIXED(Data Inspector) PCA Loadings plot not available if samples have < 30 data points.
• FIXED(Classification) Error when class names differ by leading/trailing whitespaces only.
• FIXED(Classification) Identified Class vs. X plot not rendered when feature X contains missing values.
• FIXED(Classification Setup) The number of training/test/ignore samples may be displayed incorrectly.
• FIXED(Classification Setup) Cross validation is disabled even if provided data is sufficient.

5.7, 2022-05-02

• IMPROVEDWhen editing Hard Model Peaks manually with keyboard or mouse, invalid parameter values are prevented.
• IMPROVED(MCR-ALS) Improved performance using component constraints (known concentrations or known spectra).
• CHANGEDJCAMP reader dropped support for data blocks of type PEAKTABLE.
• CHANGED(Calibration Report) The table report is now exportable.
• CHANGED(Classification Report) The table reports are now exportable.
• CHANGED(MCR-ALS Setup) The "per group"-option for the closure constraint was removed; it is unnecessary/has no effect.
• FIXEDError loading data tables when first column contains values starting with "#".
• FIXED(Classification Report) Incorrect display of the Identified Class plot when samples are plotted against any feature other than "Sample Index".
• FIXED(Classification Report) Incorrect display of the Identified Class plot when samples are colored by a feature with missing values.
• FIXED(MCR-ALS Setup) Sometimes settings are not stored for later.
• FIXED(MCR-ALS) Component spectra are scaled after (not before) checking stop criteria.
• FIXED(MCR-ALS) Error when using the Selection Tool to remove a sample which is used as "known spectrum".
• FIXED(Installer) MATLAB Runtime may be detected as missing, although it is present.

5.6, 2021-12-03

• NEWAdded 3 new reports to the MCR Component Number Analysis.
• IMPROVEDMCR-ALS Component Analysis now uses better initialization, groupwise closure constraints, and better scaling.
• CHANGEDBaseline subtraction is now turned off for the default NIR and UV-Vis session profile.
• FIXEDComponent Fitting stops prematurely in rare cases [7].
• FIXED(AppServer) The Sample() class constructor does not validate feature values.
• FIXED(Validation Report) The "Group" dropdown list does not always include the entry "All".
• FIXEDError using HMFA with distinctive peaks.
• FIXED(Calibration Report) Error when comparing calibration alternatives where some are univariate and some are PLS.
• FIXEDWrong timestamp read from Kaiser-specific SPC files on Korean or Chinese systems.
• FIXED(Identification Report) Error with Classification Models trained for a single class.
• FIXEDSorting samples behaves differently in the Main Window than in the Data Inspector.
• FIXED(Data Inspector) Profile Cursor preview doesn't get updated on changing data pretreatments.
• FIXEDWhen a session with outdated models is loaded, the dialog asking to load newer models may have no effect.

5.5, 2021-08-03

• NEW(Classification Report) Create and compare alternative classifications.
• NEW(AppServer) .NET API now supports creating NMR samples with complex y-values.
• IMPROVEDThe UI now supports the display and input of non-latin characters.
• IMPROVEDSPC file reader now supports Tornado-specific and Mettler-specific timestamps.
• IMPROVEDMagritek 1D file reader now supports reading timestamps.
• IMPROVEDDrag & Drop of CSV files now better distinguishes between Sample files and Data Table files.
• IMPROVEDRelocating data files now works for whole directory trees.
• IMPROVED(Pretreatment Model) Alignment by Reference Peak is more precise now for low-res spectra.
• IMPROVED(Calibration Model) Integration Model Components and Hard Model Components can now be linked to more than one feature.
• IMPROVED(Data Inspector) Improved colormaps for color vision deficiency.
• IMPROVED(Data Inspector) Improved editor for table filters.
• IMPROVED(Data Inspector) Improved table performance when editing/sorting/filtering.
• IMPROVED(Data Inspector) Cluster Plot (Dendrogram) now supports Selection and Data Cursor tools.
• IMPROVED(Data Inspector) Profile Cursors now keep their vertical position when the plot is scaled vertically.
• NEW(Data Inspector) The Join Table Dialog has a new option "case-insensitive match" for categorical features.
• FIXEDPLS calibration calculates wrong RMSEP and biasP for each component k (k>=2) that has different missing values than component 1 [6].
• FIXEDError reading JCAMP files with multi-line titles.
• FIXED(Data Inspector) The "Loadings on PC" slider may display numbers in a wrong format.
• FIXED(Data Inspector) Switching to the Plotter Tab loads all samples instead of filtered samples.
• FIXED(Data Inspector) Error on copy & paste of invalid values into the Timestamp column.
• FIXED(Data Inspector) Wrong coloring of PCA Scores Plot if the x-feature has missing values.
• FIXED(Calibration Setup) Error when ratiometric regression is selected and all samples have a missing feature value.
• FIXEDIncorrect handling of timestamps before the year 1582 (Julian calendar).
• FIXED(AppServer) The Analyzer.Models collection may contain duplicate entries.
• FIXED(AppServer) AddModel() error when adding a Hard Model whose components are all empty.
• FIXED(AppServer) PerformPeakPicking() error for samples with missing #ID in their URI.
• FIXEDChanging the colormap for error bar plots does not change the line color.
• FIXED(Validation Report) The Selection Tool is missing the "Select in Main Window"-option.
• FIXEDRare error when exporting to clipboard.
• FIXEDRare error during mouse-modification of model components.

5.4, 2020-11-20

• NEW(AppServer) Added AnalysisType.Auto which lets the model dictate the analysis.
• NEW(AppServer) Added "FullName" to AnalysisResult objects.
• IMPROVED(Data Inspector) Editing large tables is significantly faster now.
• IMPROVEDJCAMP reader is significantly faster now.
• IMPROVEDJCAMP reader now supports files with multiple data blocks.
• IMPROVEDImproved peak fitting, especially in case of many peaks.
• IMPROVEDCalibration of Hard Models is faster now; unnecessary component fits get skipped.
• CHANGED(Data Inspector) The 2D Samples Plot now displays samples in gray when no c-feature (color) is available.
• FIXEDSPC reader fails to find data in single-spectrum files.
• FIXEDSPC reader may retun incorrect timestamps from iCRaman files.
• FIXEDWhen a lot of samples are selected in the Samples Panel, loading additional samples becomes extremely slow.
• FIXEDLicense Activation Dialog sometimes opens in the background.
• FIXEDWhen the License Activation Dialog gets opened from the PEAXACT main window, no messages are shown when the checkout of a network license fails.
• FIXEDWhen zooming, sometimes the plot does not get drawn correctly.
• FIXEDNetwork licenses do not get released on shutdown of the PEAXACT App.
• FIXED(AppServer) Custom analysis returns sub-results (e.g. prediction uncertainties) even if not defined in the Custom Model.
• FIXEDProgress window not closing on "Share Session > Offline"
• FIXEDError when pressing [Windows key] + [Right arrow key] while no samples are selected in the Samples Panel.
• FIXEDError on startup if a per-machine license was activated.
• FIXED(Data Inspector) Changing some pretreatments shows a progress window three times.

5.3, 2020-08-20

• NEWA welcome screen is shown when starting PEAXACT for the first time.
• NEWMCR-ALS can now use a grouping feature for groupwise constraints.
• NEW(Data Inspector): CTRL+V now repeates values to fill selected table cells.
• NEWThe License Center now enables fully automated license transfer.
• CHANGEDColormap "Jet" has been replaced by the perceptually uniform colormap "Turbo".
• CHANGEDThe "Update available"-dialog has been removed. The splash screen now displays a web link when updates are available.
• FIXEDFailed to read data from JCAMP-DX files of type XYPOINTS.
• FIXEDError when reading data from Excel files when the first column contains dates.
• FIXEDInstaller sets MCR_CACHE_ROOT to a user-specific path instead to a common path.
• FIXEDCopy-to-clipboard creates a report without line breaks.
• FIXEDLicense Activation Dialog crashes in case of errors during post-activation (e.g. when saving files to a directory without write permissions).
• FIXEDMissing text output to command window.
• FIXEDError when trying to perform MCR or HMFA analyses.
• FIXED(Data Inspector) The Join Table operation incorrectly adds ignored features.

5.2, 2020-06-01

• NEWNew NMR demo data and tutorial.
• NEWThe Selection Tool now has a context menu option "Select in Main Window".
• IMPROVEDPLS models created prior to v5 now also provide prediction uncertainties.
• IMPROVEDPLS models created prior to v4 now also provide RMS residuals and Mahalanobis distance statistics.
• IMPROVEDSignificantly improved performance when evaluating Custom Results.
• FIXED[CRITICAL] Incorrect mean-centering/scaling of spectra for PLS models [5]
• FIXEDUnable to place data cursors on some plots in the main window.
• FIXEDError when clicking very rapidly while using the DataFilter tool.
• FIXEDIncorrect scaling of error bar charts when bars are disabled.
• FIXEDError when trying to load a session file if the Data Inspector has unsaved changes.
• FIXEDRare error when loading certain CSV files and Data Table files.

5.1, 2020-04-09

• NEW(Desktop App) Added a user preference for showing/hidding error bars
• NEW(ProcessLink) Added support of categorical results from Custom Models
• NEW(ProcessLink) Added support for sample identification using Classification Models
• FIXED[CRITICAL] Fitting Mode "Maximal" ignores the Component Shift Mode [3]
• FIXED[CRITICAL] Incorrect smoothing of low-end points of a spectrum [4]
• FIXEDSome toolbar and menu item may be visible although they have no effect depending on the license
• FIXEDError on classification when the Calibration Report Window was visible before
• FIXEDError when trying to create the Hard Model XY-Data report
• FIXEDModel Details Report is not formatted correctly
• FIXED(Data Inspector) Changing pretreatments when the Samples (3D) plot is visible does not update the plot correctly
• FIXED(Data Inspector) Plots are rendered incorrectly when all samples have Quality = bad
• FIXED(Calibration Report) Changing the Usage of samples may not update plots
• FIXED(AppServer) Added the missing Traverse() extension method
• CHANGED(AppServer) Removed dependency to NInject.dll

5.0, 2019-12-02

• NEWSupport for categorical (text) features.
• NEWClassification and identification of categorical features using the new Classification Model.
• NEWUser-definable results with custom mathematical expressions using the new Custom Model.
• NEW(Data Inspector) Join features from multiple tables into one.
• NEW(Data Inspector) PCA Loadings Plot.
• NEW(Data Inspector) Dendrogram Plot (Cluster Plot).
• NEW(Data Inspector) Correlation Plot.
• NEW(Pretreatment Model) Transform NMR FIDs to spectra.
• NEW(Pretreatment Model) NMR-specifc pretreatments: apodization, zero-filling, phase correction.
• NEW(Pretreatment Model) Alignment of the x-axis at a reference peak.
• NEW(Pretreatment Model) Normalize by peak area, using a configurable peak.
• NEW(Pretreatment Model) Data Filters for multiple purposes, e.g. visually highlighting spectral regions, PLS calibration.
• CHANGED[CRITICAL] (Pretreatment Model) Rubber Band Baseline subtraction now also applies to excluded ranges [1].
• NEW(Integration Model) Create components with more than one peak.
• NEW(Integration Model) New baseline types with customizable baseline ranges.
• NEW(Hard Model) Remove components and release their peaks into the editable component.
• NEW(Hard Model) A new Component Fitting algorithm is introduced which is more thorough than the previous one. Models created before v5 continue to use the previous algorithm.
• CHANGED(Hard Model) The plot of the first component now is always visible.
• CHANGED(Hard Model) Simplified Component Shift handling during Component Fitting.
• CHANGEDThe meaning of feature "Usage" has changed. It is now only relevant for calibration.
• NEW(Calibration Model) Calibration with missing values.
• NEW(Calibration Model) PLS calibration with component-specific Data Filters.
• NEW(Calibration Report) The calibration report window now also displays samples tagged as "ignore".
• NEW(Calibration Report) The calibration report window now supports the creation and editing of alternative calibrations, including removal of samples, changing sample usage between train/test/ignore, and editing Data Filters interactively.
• NEW(Analysis) Identification report using the new Classification Model.
• NEW(Analysis) Custom Analysis report using the new Custom Model.
• NEW(Analysis) Prediction report now provides prediction uncertainties (95% level of confidence).
• NEW(Analysis) MCR can be configured to use known concentrations and known spectra.
• NEWA new sample feature "Quality" was added to tag samples as good or bad.
• NEWNew selection tool for highlighting points and curves of the Data Inspector and Result Windows. The selection is kept across different reports.
• IMPROVEDNew mouse behavior for selecting, zooming, panning.
• NEWReport Windows provide different colormaps for line plots.
• NEWSupport for Specwin32 SPV-files.
• NEWSupport for Renishaw WDF-files.
• NEWSimplified license activation by downloading a purchased license from the S-PACT web server.
• NEWUser profiles are now attached to sessions. Saving/loading a session now also saves/restores the attached user profile.
• NEWSamples stored in the calibration model/classification model can now be exported to a data table.
• NEW(AppServer) New API to account for changes introduced in PEAXACT 5.
• CHANGED(AppServer) The previous .NET API still exists as a wrapper around the new API.
• CHANGED(AppServer) The previous COM API has been replaced by a new COM API.
• CHANGED[CRITICAL] The Magritek 1D file reader has changed [2].

Upgrade from v4 and Older Versions

• PEAXACT 5 is 64-bit software and requires a 64-bit Windows operating system. PEAXACT 5 cannot be installed on 32-bit versions of Windows.
• PEAXACT Application Server has been renamed to PEAXACT AppServer. PEAXACT AppServer 5 still supports both 32-bit and 64-bit applications.
• PEAXACT 5 is based on a new version of the MATLAB Runtime. Check the new system requirements and follow the installation instructions.
• PEAXACT 5 can be installed side-by-side with PEAXACT 4 or 3. The PEAXACT 5 Installer does not upgrade existing installations of PEAXACT 4 or 3. If you decide to remove earlier versions (see next point) you should first uninstall them BEFORE installing version 5.
• You are good to uninstall PEAXACT 4 or 3 except for the following scenario: If you need to save model files compatible with PEAXACT 4 or 3, e.g. when you are running the PEAXACT Application Server 4 on a computer in your production process which cannot be upgraded presently, you should keep PEAXACT 4 installed.
• If you are using the HoloPro Custom Interface it will access the latest installed version of the PEAXACT AppServer. You should uninstall PEAXACT 4 or 3 before installing version 5.
• If you are using the OPUS Custom Interface you need to upgrade your OPUS scenario scripts to use PEAXACT AppServer 5 (see PEAXACT AppServer 5 user manual).
• PEAXACT 5 requires a new license file.
• PEAXACT ProcessLink is now an independent product (instead of an extension to AppServer) and requires separate license activation.
• License modules "PI Calibration" and "HM Calibration" have been merged to "Univariate Calibration Builder".

Compatibility Considerations

• PEAXACT 5 is backward compatible. You can load model and session files saved with earlier versions. Model and session files saved with PEAXACT 5 cannot be loaded by earlier versions though.
• You may notice minor numerical differences when comparing model results of version 5 with results of version 4 or 3 due to the new MATLAB Runtime.
• [1]Rubber Band Baseline subtraction now also applies to excluded ranges, causing different results in combination with Peak Normalization when the reference peak is within an excluded region. If you created models with version 4.x, and your model uses a combination of (1) rubber band baseline subtraction, (2) excluded ranges, and (3) peak normalization, and the x-region of the reference peak for peak normalization lies within any of the excluded ranges, then you need to re-validate the model in PEAXACT 5.
• [2]The behavior of file reader for Magritek 1D files has changed. Instead of reading real and imaginary parts as 2 separate samples, it now reads 1 sample containing complex numbers. Instead of sample IDs "Re" and "Im", it now uses ID "1". If you saved a session file with PEAXACT 4 containing references to old sample IDs "Re" or "Im" you can still load the session file with PEAXACT 5, but you need to unload all 1D files from the session and then reload the files so that the sample ID gets updated. If you created Data Table files with PEAXACT 4 containing references to old sample IDs, you need to upgrade the Data Table file.
• [3]The Hard Model Fitting Mode "Maximal" now correctly honors the Component Shift Mode. Models created with PEAXACT v5.0 using (1) Fitting Mode "Maximal" and (2) Component Shift "no" behave differently now and should be re-validated.
• [4]The M=1/2*([Smoothing Filter]-1) points at the lower end of a spectrum are now smoothed correctly. Models created prior to PEAXACT v5.1 using (1) Smoothing/Derivatives with a large Smoothing Filter and (2) not excluding the M low-end points of the spectrum (e.g. by means of a Global Range or Excluded Ranges) behave slightly different now and should be re-validated.
• [5]Mean-centering and scaling of spectra as a pretreatment step to PLS calibration is now performed correctly when the calibration matrix contains missing values. Models created with PEAXACT v5.0 or v5.1 using (1) a PLS calibration, (2) with more than 1 calibrated component, and (3) calibrated with missing values (e.g. some training samples only provide feature values for component A and others only for component B) behave differently now and should be re-validated.
• [6]Calibrations made with PEAXACT v5.0 to v5.4 using (1) a PLS calibration, (2) more than 1 calibrated component, and (3) a calibration matrix with different missing values for components 2 to N than for component 1, may have resulted in incorrectly calculated RMSEP/biasP. The fewer samples were used in total, and the fewer samples were "shared" by multiple components (i.e. the more missing values the calibration matrix had), the more pronounced the deviation from the correct RMSEP/biasP value was. Note that predicted values always have been calculated correctly! Only RMSEP/biasP were affected, which may have influenced your decision process during calibration.
• [7]Component Fitting stops prematurely when the Hard Model (1) has peaks with a Max > 2 and (2) those peaks are outside the global range or inside excluded ranges and (3) the component containing those peaks is fitted with a component weight less than max(spectral intensity)*(Component Weight Threshold). If all three conditions apply, models created before v5.6 behave differently now and should be re-validated.
• [8]RMS residuals of Hard Model Fits may get calculated too large for models (1) with a Hard Model, (2) created with v5.0 to v5.7, (3) where Component Shift = off, (4) having Hard Model Peaks with a height that is orders of magnitude (specifically: (1/component weight threshold)) greater than a to-be-fitted spectrum (after pretreatments). If, in addition, such a model is calibrated, then the calculation of the RMS Residuals Outlier Probability is also incorrect and the model must be recalibrated. The problem does not occur for models without Hard Model, or created before v5.0 / after v5.7, or with Component Shift = on, or if Hard Model Peaks are in the same order of magnitude as (or smaller than) the spectra to be fitted.
• [9]Incorrect NMR preprocessing for models (1) created with v5.0 to v5.7, (2) having Transformation Mode = Spectrum Processing and Phase Correction not None, (3) for NMR spectra with complex y-values, and (4) where all imaginary y-values are zero. Such models behave differently as of v5.8 and must be revalidated.
• [10]Incorrect calculation of outlier statistics (RMS Spectral Residuals, Mahalanobis Distances, and correpsonding outlier probabilities) for groupwise predictions, i.e., predictions made with (1) a groupwise Calibration Model for (2) samples with matching grouping feature. Note that calculations of non-outlier statistics, or made with a non-groupwise Calibration Model, or for samples with non-matching grouping feature were always correct.