NEWS

---

robmed 1.1.0

• Changed default bootstrap confidence intervals to simple percentile intervals.
• Updated vignette and examples in help files.
• Fixed typos in help files.

robmed 1.0.2 (2023-06-16)

• Fixed typos and references in help files.
• Updated package description.
• Bugfix in test_mediation() and retest(): If the number of bootstrap samples is smaller than the number of observations, then the acceleration constant of BCa confidence intervals cannot be computed with (the default behavior of) boot.ci(). Now percentile confidence intervals are computed in such a case (with a warning if the user explicitly asked for BCa intervals).
• test_mediation() and retest() now give a warning if a confidence level outside the interval [0,1] is supplied.

robmed 1.0.1 (2023-03-25)

• Minor change in warning message regarding contrasts of indirect effects.
• Fixed unit tests for the case that ATLAS is used as BLAS.

robmed 1.0.0 (2022-08-17)

• Updated documentation to include reference to article in the Journal of Statistical Software.
• Updated references in help files and package vignette.

robmed 0.11.0 (2022-05-23)

• Added function sim_mediation() and wrapper rmediation() to generate data from a fitted mediation model.
• Updated documentation of fit_mediation() and test_mediation() with a brief description of the implemented mediation models.
• Updated and added more examples in help files.
• Updated the help file of the BSG2014 data with more details on the measurement scales of the variables.
• Updated package vignette.

robmed 0.10.1 (2022-03-17)

• Updated package vignette.

robmed 0.10.0 (2022-02-21)

• Added package vignette.
• Added support for the serial multiple mediator model in bootstrap tests based on regressions.
• Moved shiny app in separate package 'robmedShiny': https://github.com/aalfons/robmedShiny
• Updated BSG2014 data with more variables.
• Updated examples.
• Removed superfluous package MASS from Suggests.

robmed 0.9.0 (2021-06-04)

• Updated package dependencies.
• retest() now uses values of original object as defaults for arguments that are not supplied.
• Added argument 'contrast' to fit_mediation() and test_mediation(), which allows to test for differences of (absolute values) of indirect effects.
• Bugfix in plots: order of different models, effects or methods is preserved with R >= 4.0.0 (where data.frame() by default no longer converts character strings to factors).
• Added support for multiple independent variables to be mediated in bootstrap tests based on regressions.
• Removed demos from package.

robmed 0.8.0 (2021-01-06)

• Changes in class structure regarding how effects are stored.
• Added more unit tests.
• Bugfix in setup_density_plot(): in case of multiple indirect effects, the column "Effect" of the data frames is now a factor to preserve the order in the plot.
• Bugfix in setup_ellipse_plot() regarding checks of variables to be plotted on vertical and horizontal axis.
• Added functions setup_weight_plot() and weight_plot() to produce a diagnostic plot of robust regression weights.
• summary() method for robust boostrap tests with MM-regressions now by default produces a diagnostic plot of robust regresssion weights.

robmed 0.7.0 (2020-06-08)

• Removed deprecated functions and arguments.
• fit_mediation() and test_mediation() now allow to fit regression models with skew-normal, t or skew-t error distribtions.
• p_value() now offers functionality in line with coef() and confint(), but default behavior is kept for back-compatibility with previous versions.
• setup_ci_plot() and ci_plot() now allow to include information on p-values in addition to confidence intervals.

robmed 0.6.0 (2020-02-13)

• Added formula interface.
• fit_mediation() now checks for variable type rather than converting all variables to numeric.
• fit_mediation() and test_mediation() now use 'robust = "median"' instead of argument 'median'.
• Renamed argument 'other' for type of confidence intervals to 'type' and also changed possible values to match argument for extracting coefficients.
• New plot function ellipse_plot() for a diagnostic plot with a tolerance ellipse.
• dot_plot() has been renamed to ci_plot() and is now creating panels for different effects instead of different methods.
• density_plot() now uses argument 'fill' only for the filled rectangle that illustrates the confidence interval, and allows to customize the transparency of the confidence interval via argument 'alpha'.
• Bugfix in density_plot(): avoids that duplicated or unknown arguments are passed to geom()'s via the ... argument.
• Functions ci_plot() and density_plot() are now exported and replace plot_mediation().
• New functions setup_ci_plot(), setup_density_plot() and setup_ellipse_plot() to extract relevant information for plots.
• fortify() and plot_mediation() are now deprecated.
• Updated demos for empirical cases 2 and 3.

robmed 0.5.0 (2019-10-22)

• Changed components 'c' and 'c_prime' to 'direct' and 'total', respectively, to avoid confusion with different notation in different papers.
• Removed deprecated functions.

robmed 0.4.0 (2019-09-13)

• Output of summaries restructured to be similar to PROCESS macro of Andrew Hayes.
• Wrapper function indirect() is now deprecated.
• Added missing reference in help file of example data.

robmed 0.3.0 (2018-11-05)

• Added method based on median regressions (Yuan & MacKinnon, 2014).

robmed 0.2.0 (2018-09-25)

• Added shiny app for comparing bootstrap methods on simulated data with outliers.
• Added support for multiple mediators for standard and robust bootstrap tests based on regressions.
• Redesigned functions to live in the tidyverse: using under_scores instead of camelCase; first argument of fit_mediation() and test_mediation() is a data frame.
• Added method based on robust regressions and fast and robust bootstrap, which is the new default method.
• Wrapper function indirect() now always uses the bootstrap test

robmed 0.1.1 (2016-09-14)

• Bugfix in density plot: bootstrap replicates for indirect effect are now correctly extracted.