Release Notes Version 7 8: Difference between revisions

From Eigenvector Research Documentation Wiki
Jump to navigation Jump to search
imported>Jeremy
imported>Jeremy
No edit summary
Line 13: Line 13:
* Always reconcile variables in test set with model, even when the NUMBER of variables matches
* Always reconcile variables in test set with model, even when the NUMBER of variables matches
* Add support for "wide" (thick) venetian blinds
* Add support for "wide" (thick) venetian blinds
* Give advice when current cross-validation conditions are non-optimal (in Cross-validation settings window AND in Analysis window)
* Give advice when current cross-validation conditions are non-optimal
* Improve cross-validation speed
* Improve cross-validation speed
* ANN: fix for bug which showed the wrong CV predicted values for models with > 1 Node (always shows the CV Pred for 1 node)
* ANN: fix for bug which showed the wrong CV predicted values for models with > 1 Node (always shows the CV Pred for 1 node)
Line 22: Line 22:
* Speed improvements for iPLS, iPLSDA, Stepwise MLR
* Speed improvements for iPLS, iPLSDA, Stepwise MLR
* Calculate and show area under the curve (AUC) for ROC curves
* Calculate and show area under the curve (AUC) for ROC curves
* Calculate and display the min/max/ave percentage of samples left out per cross-validation cycle
* Add ability to automatically create plots when creating a report
* For two-class PLSDA models, include the average number of samples CORRECT as additional column (translation of fractional error into actual number of samples)
* Better exporting of text-based information in Report Writer (e.g. Confusion tables and matrices)


===Preprocessing===
===Preprocessing===
* Log10: Use an offset AND ignore negative values
* Log10: Use an offset AND ignore negative values
* Normalize: Allow normalization with multi-way data
* Normalize: Allow normalization with multi-way data
* Arithmetic: Add visual variable selection
* Arithmetic: Add visual variable selection and clarify other options
* GLSW: Make sure data .include field is applied to external data
 


===Browse===
===Browse===
Line 32: Line 38:
* Add new drag drop behavior for datasets onto datasets, datasets onto preprocessing, and datasets onto models.
* Add new drag drop behavior for datasets onto datasets, datasets onto preprocessing, and datasets onto models.
* Add support for F2 = rename selected variable
* Add support for F2 = rename selected variable
* Easier to navigate EigenGuide video list and improved handling when new list can't be obtained
* Give option to clear data from base workspace if large data is pushed into Analysis or Image Manager
* Give option to clear data from base workspace if large data is pushed into Analysis or Image Manager


Line 39: Line 46:


===Importers===
===Importers===
* Add new Bruker XRPD importer (.raw file format)
* Add new Bruker XRPD importer (.raw file format)
* Improved handling of imports when multiple importers match the file type
* Text graphical import tool - improved support for ignoring data, automatically detecting types, and visibility
* Text graphical import tool - improved support for ignoring data, automatically detecting types, and visibility
* Streamlined importing from GUIs- no longer necessary to select file type first, simply locate and load the file(s) of interest
* Streamlined importing from GUIs- no longer necessary to select file type first, simply locate and load the file(s) of interest

Revision as of 16:42, 23 May 2014

New Features in Solo and PLS_Toolbox

ASCA (Analysis of Variance Simultaneous Component Analysis)

  • Add new command-line and Analysis Window-based access to ASCA method
  • Allows multivariate analysis of designed experiments by creating a PCA model for each of the ANOVA-calculated effect matrices.
  • In the Analysis window, user can navigate through each factor's PCA model investigating the magnitude and influenced variables.
  • Effect of each factor are reported in the Effects column and in the scores for the PCA model for the particular factor.
  • The variable(s) which are significantly influenced by the factor are given by the loadings for the PCA model on that factor.
  • Includes a permutation calculator which, when enabled, will report the probability that the given effect is significant (based on randomization of samples within each level of each given factor)

Analysis Interface

  • Add ASCA (ANOVA Simultaneous Component Analysis) analysis method
  • Always reconcile variables in test set with model, even when the NUMBER of variables matches
  • Add support for "wide" (thick) venetian blinds
  • Give advice when current cross-validation conditions are non-optimal
  • Improve cross-validation speed
  • ANN: fix for bug which showed the wrong CV predicted values for models with > 1 Node (always shows the CV Pred for 1 node)
  • Add showyinbiplot option which allows hiding y-loadings in the Biplot
  • Add sorting option for dates in cache.
  • Fix for pointer not always getting set back to arrow
  • Add "weights" option to PLS and PLSDA to weight samples
  • Speed improvements for iPLS, iPLSDA, Stepwise MLR
  • Calculate and show area under the curve (AUC) for ROC curves
  • Calculate and display the min/max/ave percentage of samples left out per cross-validation cycle
  • Add ability to automatically create plots when creating a report
  • For two-class PLSDA models, include the average number of samples CORRECT as additional column (translation of fractional error into actual number of samples)
  • Better exporting of text-based information in Report Writer (e.g. Confusion tables and matrices)

Preprocessing

  • Log10: Use an offset AND ignore negative values
  • Normalize: Allow normalization with multi-way data
  • Arithmetic: Add visual variable selection and clarify other options
  • GLSW: Make sure data .include field is applied to external data


Browse

  • Align variables (if possible) when joining two datasets in the samples dimension
  • Add new drag drop behavior for datasets onto datasets, datasets onto preprocessing, and datasets onto models.
  • Add support for F2 = rename selected variable
  • Easier to navigate EigenGuide video list and improved handling when new list can't be obtained
  • Give option to clear data from base workspace if large data is pushed into Analysis or Image Manager

Dataset Editor

  • Add View > Class and Axisscale Summary menu to EditDS (views Class Summary table)
  • Add View Boxplot to menus (simple boxplot call)

Importers

  • Add new Bruker XRPD importer (.raw file format)
  • Improved handling of imports when multiple importers match the file type
  • Text graphical import tool - improved support for ignoring data, automatically detecting types, and visibility
  • Streamlined importing from GUIs- no longer necessary to select file type first, simply locate and load the file(s) of interest
  • Automatically sort filenames into alphabetical order when importing
  • Horiba importer (hjyreadr) automatically transpose image so that it matches Horiba's image mode
  • Text importer (textreadr):
    • Date format yyyymmddTHHMMSS (datestr format 30) is now recognized automatically
    • Improved handling of small data tables
    • Fix handling of double-quoted label that looks numeric
    • Recognize dates which use spelled out month names
    • Add improved auto-detection of XLS (look for binary file)
  • SPC importer (spcreadr):
    • Fix for bug reading files with negative scaling exponents (fexp<0)
    • Fix to better handle files that have binary logs as well as text audit logs
    • Add ability to specify data is an image by including "imagesize=..." and "imageorder=..." in audit log
    • Add ability to specify author and name to be used by including them as key=value pairs in audit log

Plot Controls

  • Add "automonotonic" flag to plotgui to avoid use of monotonic in "auto" mode unless allowed
  • Add scroll-wheel functionality (zoom plot, combine with shift or control to zoom one axes, click and move to pan)
  • Add density-plot replot on zoom functionality
  • Add click-to-highlight feature to move a line or class set to the front of the plot
  • Significant improvement to speed of updating when plot control listbox does not need to be repopulated
  • Add logic to bring target figure back to focus after docking controls (avoids issue that clicking on figure looses focus on figure)
  • Allow ALL options to be seen in settings
  • Add class connect methods "spider" and "means" (connect means of each x)
  • Change connect class toolbar button into a popup menu
  • Convert Axis Settings from instant-pop-up to context-menu-based command
  • Axis Settings "all fonts" option now changes axes fonts too
  • Change class colors base to improve visibility
  • Move axis lines and class outlines to bottom of plot order so that right-click controls are not inaccessible
  • Exporting figure: add new "resizemode" option to control resizing behavior
  • Revise classmarkers to improve speed and automatically tie colors to classcolors

Hierarchical Model Builder

  • Add ability to set preferences including whether Q and T2 tests are based on reduced statistics and what thresholds should be used to trigger "otherwise" in PLSDA-based rule nodes
  • Add scores as ouptut option for regression models
  • Add "on-canvas" display of numerics, strings, errors, and model types
  • Added ability to set condition on end points before the content of the end point has been defined
  • Improve behavior when overwriting nodes with similar node type
  • Allow drag/drop of models onto nodes to overwrite existing rule node WITHOUT changing branches (if possible)
  • Allow double-click of model-based rule nodes to open model in Analysis
  • Fix for error using "load" toolbar button or menu option
  • Add "copy image to clipboard" functionality.

Model Optimizer

  • Fix bug that causes error when no y-block is loaded.

mpca / parafac

  • Allow for single slab input for prediction if sizes of two existing modes match variables for n-way array


New Command-line Features and Functions

Misc New Functions

Command-Line Changes

  • anovadoe
    • Fix to force use of center points
    • Use design information to identify center points ONLY (don't infer based on values)
    • Allow input of matrix for y (squelches statistical tests - only reports coefficients)
    • Add calculation of 3-way interactions
    • Calculate the ANOVA decomposition column/matrices as needed for ASCA
  • boxplot
    • Fix for bug which would truncate labels to a single character
    • Fix for bug when converting multiple inputs to options structure when one of those inputs was a cell array
    • Revision to logic - give preference to classes in DSO when grouping (if no classes passed) THEN use labels (if no classes present in DSO)
  • brukerxrpdreadr -add new Bruker XRPD importer (.raw file format)
  • confusionmatrix / confusiontable
    • Handle excluded samples in simca case.
    • Allow cell array of strings to be passed as trueClass and predClass
    • Fix for bug in texttable output when no y is present
  • dataset
    • -Add augment method, will create class that tracks augmented dataSets after a concatenation
    • -Add table method, gets text table with labels from DataSet
  • Experiment Designer -Add randomize menu item, overhaul options usage, and fix a few bugs with loading a DOE dataset.
  • oplecorr -Add new "Optical Path-Length Estimation and Correction" preprocessing (command-line only)
  • textreadr -Add "catdim" and "autopermute" options to manually control concatenation of multiple files