jasp-stats/jasp-desktop v0.18.2

Release 0.18.2

on GitHub

New Features and Improvements

• New module:

  • Survival (Beta)
  • Process (Beta)

• Data Editing:

  • Redesign the Variable setting window
  • Possibility to switch computed columns to not computed and vice-versa
  • Add empty values settings per variable
  • Make JASP readable on small screen
  • Copy/Paste with headers work between JASP and other spreadsheet editor.
  • Undo/Redo shortcut available everywhere on data editing mode

• Results:

  • Add video support in annotation (#5293)

• Windows Installation

  • Add MSIX Installation in Windows Store.

• Audit:

  • implements algorithm auditing (i.e., model fairness, a new section in the module): jasp-stats/jaspAudit#371

• Factor/CFA

  • Option to fix intercepts to zero in mean structure (jasp-stats/jasp-issues#2223)
  • Add structural invariance and mean structure identification options (jasp-stats/jasp-issues#2378, jasp-stats/jaspFactor#193)

• Machine Learning:

  • Add Naive Bayes classification analysis (jasp-stats/jaspMachineLearning#244):
    Implements some explainable modelling ('XAI') features from the DALEX package
    • After model performance, implemented an option in all supervised analyses to show feature importance metrics (via https://ema.drwhy.ai/featureImportance.html).
    • After feature importance, implemented an option in all supervised analyses (and the prediction analysis) to explain the predictions of the model as a sum of feature contributions (via https://ema.drwhy.ai/breakDown.html#breakDown).
    • Added multivariate normality check in the LDA analysis (https://github.com/jasp-statsjasp-issues#2272)
    • Adds a bare-bones linear regression analysis as a baseline method in which the user can use a training and test set to compare the results with other techniques.
  • Linear Regression: Add an option to display the regression equation for (regularised) linear regression (jasp-stats/jaspMachineLearning#227)
  • Optimization for support vector machines and decision trees (jasp-stats/jaspMachineLearning#84)

• SEM:

  • Add mean structure option fixing mean intercepts to zero (jasp-stats/jaspSem#186)

Bug fixes

• Multiple labelfilters do not combine correctly (jasp-stats/jasp-issues#2376)
• Filter still active after reopen new session (jasp-stats/jasp-issues#2344)
• Export csv file from .jasp file will lose values (jasp-stats/jasp-issues#2396)
• Latex in darkmode (#5311)
• Bayesian ANOVA: Fix Single model inference (https://github.com/jasp-stats/INTERNAL-jasp/issues/2426)
• Learn Bayes: fix bracket at logBF10 (jasp-stats/jaspLearnBayes#171)
• Machine Learning:
  • Do not scale the target variable in regression anymore.
  • Fix bugs in Neural Network analyses related to activation function use (had wrong input for linear.output argument).
• Quality Control
  • Graph variation components has not been built (jasp-stats/jasp-issues#2273)
  • Percentage Process Variation Graph hadn't been built (jasp-stats/jasp-issues#2274)
  • Attributes Agreement Analysis analysis crash (jasp-stats/jasp-issues#2286)
  • Added help files (jasp-stats/jaspQualityControl#285)
  • Fixed issue that sometimes broke the RSM analysis (jasp-stats/jaspQualityControl#281). The rsm package calculated some "canonicals" that were not used in the JASP output. Setting the threshold to calcuate these to 0 resolved the issue.
  • Removed option to specify blocks again, breaks too often.
  • Removed options to specify manual terms for RSM analysis, breaks often and is not really useful without possibility of adding squared terms.
• Regression analyses:
  • Bug in German translation (jasp-stats/jaspRegression#268, jasp-stats/jasp-issues#2405)
  • Wrong values of Nagelkerke and Cox and Snell R^2 reported (jasp-stats/jasp-issues#2368)
• Reliability
  • CI for Krippendorff's alpha (jasp-stats/jasp-issues#2318)
• SEM
  • Chi-square change with addtion of second model (jasp-stats/jasp-issues#2302)