Statistical tests¶

PySofra auto-selects per-variable tests following the tableone / gtsummary conventions. You can override per variable with tests=.

Defaults¶

Variable kind	2 groups	3+ groups
Continuous	Welch's t-test	One-way ANOVA
Continuous + `nonnormal=`	Wilcoxon rank-sum	Kruskal–Wallis
Dichotomous / categorical (2×2)	Fisher's exact	—
Categorical (larger)	—	Pearson χ² (flagged sparse if any expected < 5)

import pysofra as ps
ps.available_tests()

Continuous	Categorical
`welch` / `welch_t` / `ttest` / `t`	`fisher` / `fisher_exact`
`student` / `student_t` / `equal_var_t`	`chisq` / `chi_square` / `chi2` / `pearson`
`wilcoxon` / `mannwhitney` / `mwu` / `rank_sum`
`anova` / `oneway_anova`
`kruskal` / `kruskal_wallis`

Override per variable:

ps.tbl_one(df, by='arm',
           tests={'hba1c': 'wilcoxon', 'race': 'fisher'}).add_p()

The footnote automatically names both the overrides and any defaults that fired elsewhere.

.add_q(method='fdr_bh') adds a q-value column. Methods come from statsmodels.stats.multitest.multipletests: