|
Release History
Die folgende Tabelle gibt einen Überblick zu den neuen Features, Verbesserungen und Korrekturen der letzten Versionen von Epina DataLab. Die neueste Release kann von der Downloadseite geladen werden. Bitte beachten Sie, dass Updates, die noch nicht in die letzte Major Release eingeflossen sind, über den automatischen Updatemechanismus geladen werden können (vgl. auch die Liste der neuesten Änderungen).
Release 4.200 [2021-08-08] |
New Features
Improvements and Changes
- implemented Levene's test for testing the heteroscedasticity of the univariate regression residuals
- implemented Tukey windowing in FFT
- the number of remembered loaded files has been increased to 24
- the contingency analysis tool now calculates the Yates correction of Pearson's chi2 as well
- columns and rows can now be deleted in the matrix viewer as well
- the univariate regression tool now allows to display date/time values
- replaced the forum link in the about box by the link to the DataLab Youtube channel
- PLS/DA and MBL classifier metrics can now be copied to the clipboard via the context command 'copy to clipboard'
- improved PLS/DA: cross validation is now much faster, the MCC is calculated during CV and the number of starting factors can now be set by the user
- new statistical functions in DLabPascal: Chi2OfContTab, Chi2OfContTabYates, CorrTest, PearsonCorrCoeff, Erf, ErfApprox, EstimateProbDensity, FisherExactTest, FisherTransform, FisherTransformInv, HyperGeoDistriDensity, GrubbsCriticalValues, GrubbsTest, KSPValue, KSQuantile, PerformKSNormalityTest, Prob2x2Contingency, RunsTest, RunsTestSerial
|
Release 4.100 [2021-04-08] |
New Features
Improvements and Changes
- changed colors of loadings in PCA biplots from blue to black with gray rays
- calendar tool to pick a TDateTime value via a calendar
- extended user interface of the Hierarchical Cluster Analysis window
- HCA now allows to select individual branches of the dendrogram zo assign a class to it
- HCA now provides a calculate button and starts the clustering only when clicking this button
- all diagrams now offer a Save Diagram to Disk command in the context menu to save the diagram as an image (BMP, PNG, JPG)
- improved 'Importing Complex Text Data' tool
- VIF tool can now sort the results columnwise
- new functions in DLabPascal:
BinomCoeff, BinomDistriDensity, BinomDistriIntegral, BinomDistriQuantile, CalcFishQ, Chi2DistriDensity, Chi2DistriIntegral, Chi2DistriQuantile, ConfidenceInterval, DeanDixonCriticalValues, DeanDixonTest, DurbinWatsonCrit5pct, DurbinWatsonTest, EncodeDateTime, FDistriDensity, FDistriIntegral, FDistriQuantile, IncompleteBeta, IncompleteGamma, LillieforsPValue, LillieforsQuantile, LnBeta, LnBinomCoeff, LnGamma, MedianTest, FDistriDensity, nDistriIntegral, nDistriQuantile, Perform1SampleChi2Test, Perform1SampleTTest, Perform2SampleTTest, PerformFTest, PerformLillieforsTest, PerformMannWhitneyUTest, ShapiroWilkIntegral, ShapiroWilkTest, nDistriDensity, tDistriIntegral, tDistriQuantile, UDistriDensity, UDistriIntegral , UDistriQuantile, VarInflatFactor, WilcoxonSignedRankTest
|
Release 4.000 [2021-02-05] |
New Features
- DLabPascal now provides an integrated debugger
- several extensions of the built-in DLabPascal library
- signal integration tool
- a tool for calculating binary classifier metrics
- new statistical test: median test
Improvements and Changes
- Wilcoxon test report now displays the median of the differences as well
- redesigned random forest form
- a variable set repository makes the selection of variables much more convenient
- the RF applicator has now an option to create binary results for classifiers
- the PLS Applicator now shows the CF thresholds and allows to store the results in the aux. maxtrix or append it to the data matrix
- the data editor now supports ^C and ^V for copy and paste (instead of 'extract from clipboard')
- refurbished k-means clustering
- RBF network extended
- PLS/DA now displays more classifier metrics (same as MBL)
- numerical editor: row names are now displayed using a much longer maximal length; two additional fonts (Tahoma, Verdana) included
- MLR and Ridge Regression diagrams provide now more plot options
- memory based learner improved in many details
- refurbished Kohonen map
- improved edit/selection mode in data description window
- additional performance metrics for RF classifiers
- formula editor now uses the ° symbol as cell reference
- class color setup dialog now supports saving and loading color palettes
- the default class color palette has been improved
|
Release 3.951 [2020-03-26] |
New Features
- Linear Discriminant Analysis (LDA)
- new statistical tests: chi2 goodness-of-fit, contingency analysis, adjusted Rand index
- extended DLabPascal:
- full support of the programmatic control of the plot parameters
- many extensions of the built-in library
- memory based learner (kNN) is now supported by DLabPascal
- univariate regression tool can now generate a DLabPascal script
Improvements and Changes
- user interface of the growing neural network improvd
- added a new and better language selection dialog
- the symbol used for dot attribute in 2d plots can now be selected by the user
- all save dialogs are now exhibiting a consistent user interface
- major revision of the memory based learner (kNN)
- added documentation for TAssocArray
|
Release 3.911 [2019-04-05] |
Bug Fixes
- the scaling tool did not work under Windows 7
- DataLab generated an access violation if the univariate statistics were open when adjusting the table size
|
Release 3.910 [2019-03-29] |
New Features
- DataLab provides now an "open windows" selector which allows to switch between windows
- the data plots now provide a graphical marking tool to apply red and blue marks using either a lasso or a rectangle
- DataLab offers now a universal scripting language called DLabPascal which allows to automate recurring tasks
- the newly implemented memory based learning tool provides powerful option both for regression modelling and classification
- a new baseline correction tool allows to correct the baseline of a signal
- all windows can now be detached from the DataLab master window to allow for multiple monitors
- DataLab now supports random forest models (both regression models and classifiers)
- contingency analysis and Fisher's exact test are now available for the analysis of nominal data
Improvements and Changes
- pressing the ESC key now closes the current window (if appropriate)
- various MLR improvements: Cook's distance is now calculated and displayed as a graph; Durbin-Watson-Test now displays the critical limits and the decision
- drawn from the test; added 'copy y-hat to matrix clipboard' to the MLR results page;
- ridge regression: implemented a trace of parameter significance
- fixed zoom mode of 2D plots improved
- implemented binomal distribution in distribution calculator
- the desktop color can now be set by the user
- the digitizer accepts now PNG images as well
- extended 2D data designer: data can be loaded from main data table; two specific columns of the data matrix can be overwritten
- the plot window now has coordinate and object display
- the scaling tool now supports q-normalisation ánd squashing functions
- various improvements of the simple regression tool: data points can now be connected, polynomial coefficients are now evaluated by adding the p-values, residuals can now be plotted against the object index, plot parameters can now be adjusted
- cluster analysis: base line is now moveable to allow for long object names
- implemented the creation of mutually exclusive test and training sets in the data split tool
Bug Fixes
- normal probability plot crashed if the data to be displayed was constant
- stepwise regression hung in certain cases
- 3D surface plot did not handle nominal and ordinal variables correctly
- adjusted R2 in MLR was wrong
- command 'Clear Matrix Clipboard' did not clear the matrices 5 to 8
- data designer did not correctly set the level of measurement when copying designed data over existing data
- digitizer did not load background image if the file path of the image changed
- simple linear regression: Durbin-Watson test was flawed
- help page assignments were partially wrong
- import CSV - the selection box for the separating character was incorrectly intialized
- loading future versions of IDT files were not processed correctly
- matrix stack did not correctly copy empty cells
- MLR created range check error if there were empty rows
- normal probability plot did not allow to mark the very first data object
- spectral diagrams did not include the zero line when displaying them
- XLS import crashed; plus various other minor bug fixes
|
Release < 3.910 |
Die Versionsgeschichte vor der Release 3.910 wurde aus Platzgründen gelöscht |
|