pg-nonparametriclisted

# PG Nonparametric Use when the outcome is ordinal, or continuous but non-normal with small n, so a rank-based test is safer than a t test or ANOVA. ## Load Read: - `../../references/supervision-gates.md` - `../../references/pingouin-api-quickref.md` - `../../references/apa-output-template.md` if writing results. ## Decision Rules - Two independent groups -> `pg.mwu(x, y)` (Mann-Whitney U). - Same participants measured twice / paired -> `pg.wilcoxon(x, y)` (signed-rank). - Three or more independent groups, one factor -> `pg.kruskal(data, dv, between)`. - Three or more repeated conditions, continuous/ordinal -> `pg.friedman(data, dv, within, subject)`. - Three or more repeated conditions, binary outcome -> `pg.cochran(data, dv, within, subject)`. - Follow a significant omnibus with `pg.pairwise_tests(..., parametric=False, padjust="holm")`. ## Required Inputs - Outcome column and its scale (ordinal or non-normal continuous). - Grouping (between) or condition (within) column. - Subject ID for paired/repeated designs. - Post hoc correction: default `holm`. - Alternative hypothesis: default `two-sided` (mwu/wilcoxon only). ## Code Patterns Mann-Whitney U (independent): ```python x = df.loc[df["group"].eq("A"), "score"] y = df.loc[df["group"].eq("B"), "score"] res = pg.mwu(x, y, alternative="two-sided").round(3) pg.print_table(res) ``` Wilcoxon signed-rank (paired): ```python res = pg.wilcoxon(df["pre"], df["post"], alternative="two-sided").round(3) pg.print_table(res