NEWS.md
t_test() - a comprehensive wrapper for t-tests that always returns Cohen’s d. Supports one-sample, independent-samples, and paired-samples t-tests with both formula and vector interfaces. This replaces the now-deprecated paired_t_test_d().run_mediation() to estimate and plot_mediation() to visualize parallel mediation models.na_when() to set values to NA based on logical conditions, and na_ifs() to replace multiple values with NA (naming choice guided by existing dplyr::na_if function)make_scale_mi() to estimate scale scores and Cronbach’s alpha after multiple-imputation at the item level. According to Gottschall, West & Enders (2012), this is one of the best ways to deal with item-level missing data. If requested, the function can bootstrap confidence intervals, using the future-package for parallel computing.pcor_matrix() to calculate partial correlation matrices after parceling out one or several variable.run_and_format() to run any code and return formatted code and output for sharing - in the style of reprex::reprex() but without creating a new session (intended for teaching or sharing code examples, rather than bug reporting).paste_() that mimics paste() but removes NA-valuesdupl_items() that returns (unique) duplicated items from vector, removing need for subsetting with duplicated()
make_scale_mi() output structure has been standardized to match make_scale(). The descriptives element now uses "cronbachs_alpha" (instead of "cron_alpha") for the reliability_method field, includes reversed as a space-separated string (instead of a data frame), and adds rev_min and rev_max fields for consistency with make_scale(). Note: "cron_alpha" is still accepted as input for backward compatibility.report_cor_table() and plot_distributions() now support survey-weighted data. Distribution plots for survey design objects properly account for sampling weights using svyhist() for discrete variables and svysmooth() for continuous variables. This ensures that distributions reflect the population rather than just the sample composition.timesaveR.round_method to control rounding behavior consistently across all formatting and rounding functions. Set to "to_even" (default, banker’s rounding) or "default" (standard R rounding). Change via options(timesaveR.round_method = "default"). This ensures all round_(), round_df(), fmt_p(), fmt_pct(), fmt_cor(), and related functions use the same rounding method.make_scale() now has a proration_cutoff argument to specify the maximum share of missing data ignored in each case. This offers a simple way to improve on casewise deletion for handling missing data, without getting scale scores based on an insufficient subset of items. [NB: This is a breaking change, earlier versions implicitly had a proration_cutoff of 1, returning scale scores if at least one item was present - which is indefensible when there is a lot of missing data.]text to descriptives returned by make_scale()
report_lm_with_std() to show R2 change for more than one pair of modelst.test()-style formula notion in pairwise_t_tests()
make_scales()
report_cor_table() now ensures correct ordering of extra columns if row_names column is includedtidy.svy_cor_matrix() to tidy survey-weighted correlation matriceswtd_cor_matrix_mi() to cor_matrix_mi() to reflect that weights are optionalfmt_p() gained a equal_sign argument that determines whether p-values that are reported precisely are prefixed with “=”dump_to_clip() now accepts objects passed directly, or through the pipe (#5)line_to_vector() now automatically returns numeric vectors when only numbers are passed, and has gained an option to retain NA values for blank entries. It also allows users to pass one or multiple separators to split at explicitly. l2v() is now an alias for line_to_vector().paired_t_test_d() is now deprecated in favor of the more comprehensive t_test() function. Users should switch to t_test(x, y, paired = TRUE) instead.report_polr_with_std() works again after broom::tidy.polr() update broke itfmt_p() now supports greater numbers of significant digits properlyfmt_p() now checks argument types, so that it does not return misleading results when characters are passed (#10)pairwise_t_tests() can now deal with missing data and labelled arguments (#9)make_scales() now correctly reacts to print_desc argumentcor_matrix() now works robustly with missing = "fiml" even when some bootstraps do not convergeround_df() and round_() now consistently use round-to-even/banker’s roundingsvy_miss_var_summary() to quickly check for missing data in survey objectspolr_std() function to run proportional-odds model with continuous predictors standardised (analogous to lm_std())dummy_code() for creating k-1 dummies with tidy namesreport_anova() to report F-tests for model comparisons (limited functionalities at present)cor_matrix()
ggplot_save()
fmt_p()
ggsave_show() now correctly takes file type from file extensionreport_lm_with_std() and report_polr_with_std() broke when gt implemented tidyselect. Fixed again.