Release 0.16.1

New Features and Improvements

• New languages added: French (incomplete)
• Repeated Measures ANOVA post hoc tests
  • Improved effect size calculation, including interactions, multiplicity correction, and confidence intervals
    • Effect sizes now computed with emmeans, which also allows effect sizes for interactions and confidence intervals (#1135, #1299, #1303).
  • Update help file to explain Levene test (#1518)
• Machine Learning
  • New analyses decision tree regression and classification
  • New analyses Support vector machine regression and classification
  • K-medoids and K-medians clustering techniques added to Neighborhood-based (formerly K-means)
  • Extra plot in k-nn regression and classification showing the relative weights as a function of the distance
  • Extra evaluation metrics for all classification analyses
• New module: Cochrane Meta-Analysis
  • Allows users to select and analyze data from Cochrane Database of Systematic Reviews
• Regression
  • Logistic Regression
    • Total percentage in confusion matrix added (#1471)
    • Multicollinearity Diagnostics added (#1515)
  • Improved the help file (#1495)
• Bain
  • An extra column showing PMPc has been added to the ANOVA, ANCOVA, SEM, and linear regression standard table outputs
• Descriptive Statistics
  • Option to hide/ show Valid and Missing cases (#1480)
• Factor Analysis
  • New option for parallel analysis: 1) can be based on FA or 2) PCA (#1319)
  • Scree plot is overhauled to be more concise. New option to turn parallel analysis results on and off in the plot
• Reliability
  • New analysis for Cohen’s & Fleiss’ kappa (#310, #1476)
  • Bayesian Unidimensional Reliability added
    • Standardized coefficients
    • Standardized single-factor loadings
    • Control over the priors
    • Median option as a point estimate
    • Fit measures for single-factor model (rmsea, LR, cfi, tli)
    • Improved help file
• Audit
  • Added an option to specify the maximum sample size
  • Fixed a bug in the Bayesian workflow where the sample size started counting at 5, sometimes resulting in sample sizes that did not match the desired input criteria
• Improvements to “About”
• Improved layout of network graphs in Network module (#1366)

Bug fixes

• Path plots in Factor module fixed (#1548, #1570, #1573)
• Factor correlation matrix showed the correlations of the estimated factor scores. Now the proper factor intercorrelations are used
• Computing Z-scores does not give an error anymore (#1507)
• Pairwise plots able to handle ordinal variables (#1523)