Release Notes Version 6 5

From Eigenvector Research Documentation Wiki
Jump to navigation Jump to search

Version 6.5 of PLS_Toolbox and Solo was released in October, 2011.

For general product information, see PLS_Toolbox Product Page. For information on Solo, see Solo Product Page.

(back to Release Notes PLS Toolbox and Solo)

NEW FEATURES

Classification Methods

  • Total flexibility in choosing class sets to model (through class groups interfaces or Edit > Options > Method Options menu)
  • Confusion table and matrix tools added
  • Unified reporting of classification results for SIMCA, KNN, PLSDA, and SVMDA including:
  • all class probabilities
  • most probable assignment
  • strict in-class assignment
  • strict multi-class assignments
  • KNN Cross-validation fully supported

Design Of Experients (DOE) Tools

  • Significant addition of tools for DOE generation and testing
  • Experiement Designer tool added to create, examine, and print run sheets for designed experiments
  • DOE Anova tool added to calculate significance of factors and interactions
  • Added support tools to: scale factors, calculate interactions, identify confused factors and interactions
  • Design generation tools included for:
  • Mixed-level Full Factorial Designs (factdes)
  • Fractional Factorial Designs (ffacdes1)
  • Face and Spherical Central Composite Designs (ccdface, ccdsphere)
  • Box-Behnken Designs (boxbehnken)

Other Method Improvements

  • Permutation testing of regression and classification models greatly improved with probabilities of model significance and plots
  • Support Vector Machine One-Class support added (svmoc; command-line only)
  • MCR/ALS contrast algorithm improved providing better handling of low signal cases.
  • MLR support for studentized residuals, T^2 and limits added
  • MODELSELECTOR trigger can now use of any classification model or a simple logical test on predictions from a regression or PCA model.

Preprocessing

  • Variable Alignment method added (using either peak alignment using registerspec or Correlation Optimized Warping (COW))
  • Correlation Optimized Warping (COW) and Dynamic Time Warping (DTW) added (from http://www.models.kvl.dk/DTW_COW )
  • Poisson scaling offset support added
  • Class centering method added for Multilevel PLS models and other clutter-correction methods
  • "Transform" options for data blocks in Analysis window. Context menu gives quick access to:
  • Batch Digestion (MPCA, Summary PCA)
  • Calibration Transfer
  • Polynomial Augmentation (Non-linear modeling)
  • Coadd Data Reduction (Down-sampling)
  • Calibration Sample Selection (redundant sample exclusion, re Westerhuis method)
  • Baseline removal improved with manual entry of baseline points and easier graphical selection.
  • GLSW improved to work with multi-class PLSDA classification models (multivariate logical y-blocks)
  • Help and technical details easy-access links added to Preprocessing Window
  • Descriptions automatically update to show current settings for adjustable methods

Import / Export

  • Omnic SPA file importer added
  • Galactic DHB file importer added (support in SPC reader)
  • 3-way data importing from multiple files or multiple Excel spreadsheets added
  • Excel multi-sheet document importing improved (better labeling and concatenation options)
  • XML format, improved importing of objects and image DataSets
  • System clipboard importing improved
  • Importer list easier to navigate
  • Autoexport tool added for easy saving of data to external formats

DataSet Editor

  • Summary statistics for all or part of the given data (View > Summary Statistics, context menu access) added
  • Access to all classes, labels, and axisscales on data table view (pull-down menus on Data tab)
  • Date stamps as axisscales improved. Dates are displayed as text dates and not datestamp values
  • Selection of data improved with easier selection full integration with plots of data
  • Data exchange between interfaces improved (File > Send To menu)
  • Quick access to commonly used operations (via toolbar buttons)

Plot Controls

  • "Stacked" plots (automatic y-scale toolbar button cycles through scaled data, stacked data, normal view)
  • "Color by" any related data or user-selected data
  • "Invert selection" toolbar button added
  • Improved access to data in table format
  • Persistent selection capability. Selection mode stays activated after a selection. (Optional - disable through settings)
  • Customizable line widths (context menu on the plot)
  • Customizable figure toolbar to choose which toolbar buttons are shown.
  • Time-based axes have improved automatic update after a zoom or pan

Analysis Window

  • Automatic data augmentation by simply importing / loading new data
  • Easier to read format for main table with improved information
  • Model Cache speed improvements

Workspace Browser

  • DOE Experiment Designer and easy Calibration Transfer model application added
  • Drag/drop of multiple items improved

Command-line Tool Changes

NEW FUNCTIONS AND FILES

  • Full Support for Matlab R2011b

Misc New Functions

autoexport - Exports a DataSet object to a file of the specified format.
classcenter - Centers classes in data to the mean of each class.
flucut - Remove scatter from fluorescence EEM data.
permutetest - Permutation testing for regression and classification models.
permuteplot - Create plot of permutation test results.
permuteprobs - Display probabilities derived from permutation testing.
poissonscale - Perform Poisson scaling with scaling offset.
spareadr - Reads Thermo Fisher SPA files.
specalignset - Settings used to modify variable alignment preprocessing.
svmoc - Support Vector Machine for one-class analysis.

Distribution Fitting Toolbox Tools

randomttest - Randomization t-test for evaluating residuals from two models.
signtest - Pairwise sign test for evaluating residuals from two models.
wilcoxon - Pairwise Wilcoxon signed rank test for evaluating residuals from two models.

Design of Analysis Tools

anovadoe - Function to perform ANOVA for 2^k factorial model X, Y data
boxbehnken - Create a Box-Behnken Design of Experiments.
ccdface - Create a Face-Centered Central Composite Design of Experiments.
ccdsphere - Create a Spherical Central Composite Design of Experiments.
doegen - Generate a Design of Experiments (DOE) DataSet object.
doegui - Design of Experiments tool.
doeeffectsplot - Main & Interaction Effects plots.
doeinteractions - Calculates interaction terms of a raw DOE matrix.
doescale - Convert coded DOE to scaled DOE or scaled back to coded.
factdes - Full factorial design of experiments.
ffacconfusion - DOECONFUSION Generates confusion table for a fractional factorial DOE.
ffacdes1 - FACTDES Fractional factorial 2-level design of experiments.

Other Changes

crossval - improved display of classification results, improved integration with model input
excludemissing - improved handling of unusual missing data patterns
fasternnls - improved performance for large numbers of factors and for low signal problems
parafac2 - improved memory efficiency and missing data support
residuallimit - add support for calculating confidence level from given Q value and a model
stdgen - tuned performance
summary - improved output and display options. Now outputs DSO or displays more useful data.
svmda - improved parameter selection for special cases
wlsbaseline - added variable weights as output (weighting used on each variable in spectrum)