Skip to main content
corrkit 0.2.0
- One computing function,
cor_compute(), replaces cor_test_matrix() and cor_cross(): y = NULL gives every pair of x (matrix), y gives the cross analysis x x y, one x and one y a single pair; covariates works in every case (partial correlation matrices are new). One result class, cor_result.
- Methods of
cor_result: print(), summary() (settings, variables, strongest pairs, content of the object), plot() and autoplot() with type = "heatmap", "forest" or "scatter", as.matrix(what = ), as.data.frame().
- New
plot_cor_forest(): each coefficient with its confidence interval.
plot_cor_pair(): choice of the fitted curve (smooth = "lm", "loess", "gam", "glm", a function or "none", plus smooth_args and se; default "lm" for Pearson and partial correlations, "loess" for Spearman and Kendall), colour-blind-safe default colours (Okabe-Ito: point_colour, line_colour, group_colours), shorter axis titles for the added-variable plot. The palette argument is removed.
tbl_cor(layout = "matrix" | "wide" | "long") replaces tbl_cor_matrix(), tbl_cor_cross() and tbl_cor_long().
plot_cor_heatmap() takes a cor_result and draws a triangle (matrix) or a rectangle (cross); plot_cor_cross() is removed.
cor_binary_effects() also accepts a cor_result.
- Removed:
cor_test_matrix(), cor_cross(), plot_cor_cross(), tbl_cor_matrix(), tbl_cor_cross(), tbl_cor_long(), partial_rank_cor().
- Plot texts are typeset: with the ggtext package (Suggests), subtitles, captions and axis titles are rendered as markdown, so that r_s, r_xy.z, e_x.z, r_rb… appear as an italic symbol with a subscript (
options(corrkit.markdown = FALSE) for plain text); the colour legend uses plotmath.
options(corrkit.unicode = FALSE) (default) writes the square root, minus sign and tau as sqrt, -, tau, so that figures can be saved with the pdf() device; TRUE keeps the symbols (PNG, or ggsave(device = cairo_pdf)).
corrkit 0.1.0
- First version, built from
03_correlation.R.
- Engines:
stats::cor.test(), ppcor::pcor.test(), psych::corr.test(), correlation::correlation() (new engine, gives access to Kendall and to the robust methods of the correlation package).
- Confidence intervals (Fisher z, via
correlation::cor_to_ci() when installed).
- New:
cor_cross(), plot_cor_pair(), cor_binary_effects(), gt tables tbl_cor_matrix(), tbl_cor_cross(), tbl_cor_long(), tbl_cor_binary(), example data simulate_cor_data().
- Three vignettes (in French).
Changes after the first site
- p-values are now raw by default (
p_adjust = "none") in cor_test_matrix(), cor_cross() and plot_cor_cross(); adjust_by is used only when an adjustment is requested. tbl_cor_long() hides the adjusted column when p-values are raw.
- New vignettes: detailed workflow, multiple-testing corrections, comparison of correlation coefficients.
- Site now 100% Quarto, built with altdoc (
altdoc/, workflow .github/workflows/altdoc.yaml); pkgdown removed. Vignettes in Quarto, with tabsets and callouts.
- Logo script in
data-raw/logo.R (hexSticker).
Back to top