NEWS.md
BREAKING CHANGES
Removes the alias expr_onesample_proptest.
The expr_template function retires effsize.df argument. Now all details need to be entered only in stats.df.
All meta-analyses are now carried out using expr_meta_random and the individual functions have been removed.
MAJOR CHANGES
All effect sizes for contingency tabs are now calculated via effectsize instead of rcompanion. This would lead to slight differences in effect sizes and their CIs but the computations will be faster. Additionally, the lower bound will never be negative and will be restricted to [0,1].
expr_contingency_tab function has been made less robust. It now fails instead of returning NULL when it is not supposed to work. This is done to be consistent with the other functions in the package which also fail instead of returning NULL.
MINOR CHANGES
Retires the vestigial stat.title argument. It was originally intended to give more info on the tests, but now the expressions themselves contain these details.
For paired ANOVA designs, partial = TRUE is recognized by effect sizes.
Retires bias.correct argument for contingency table analysis. It is rarely justifiable to use the biased version of Cramer’s V.
MINOR CHANGES
Adapts tests to changes made in the correlation package.
Subtitles for correlation tests make clear the type of statistic.
Small p-values (< 0.001) are now shown in scientific format.
MINOR CHANGES
Adapts to changes made in tidyBF package.
Re-exports correlation::correlation needed for ggstatsplot.
The expr_t_nonparametric subtitle now clarifies whether it’s a Wilcoxon test or a Mann-Whitney test.
MINOR CHANGES
Thanks to Sarah, the package has a hexsticker. :)
Confidence intervals for Spearman’s rho are computed using correlation instead of rcompanion.
All relevant functions get rid of messages argument as the functions no longer print a message when bootstrapped CIs are used.
The effect size measure for paired robust t-test is now changed to robust (trimmed-Winsorized) standardized difference similar to Cohen’s d.
BUG FIXES
0.4.0 release for expr_anova_parametric: changing conf.level doesn’t work and function defaults to 0.90 CIs (#32).MINOR CHANGES
BREAKING CHANGES
Removes the experimental corr_objects function.
All Bayes Factor related functions have now moved to the new tidyBF package and are re-exported from there.
MAJOR CHANGES
Minimum R version bumped to R 3.6.0.
Retires the internal effsize_t_parametric helper function in favor of relying functions from effectsize, which is now added as a dependency. Similarly, statsExpressions now relies on effectsize to compute effect sizes for ANOVA designs, instead of sjstats.
For parametric t-tests and ANOVAs, confidence intervals for effect sizes are estimated using the noncentrality parameter method. Centrality-based methods are deprecated.
Correlation analysis is carried out using correlation package, which is now added as a dependency.
MINOR CHANGES
corr_objects to reduce dependency load of ggstatsplot. This is an experimental function and should be avoided until it stabilizes.NEW FEATURES
expr_meta_bayes.NEW FEATURES
expr_meta_parametric, expr_meta_robust, bf_meta.BREAKING CHANGES
expr_template function now expects two dataframes: stats.df and effsize.df that contain the details needed for creating expressions instead of providing each individual values. This makes the function more friendly work with using modeling packages like broom.MINOR CHANGES
Minor tweaks to how widehat is displayed in some of the expressions.
Cramer’s V is bias-corrected by default.
MAJOR CHANGES
Removes MCMCpack from Depends.
All effect size texts now contain ^ on top to signify that these are estimates.
MINOR CHANGES
CRAN.