omicverse.metabol.anova

Contents

omicverse.metabol.anova#

omicverse.metabol.anova(adata, *, group_col='group', groups=None, method='welch_anova', layer=None)[source]#

Per-metabolite test across 3+ groups.

Parameters:

  • group_col (str (default: 'group')) – Factor column in adata.obs.

  • groups (Optional[list] (default: None)) – Subset of levels to test. None → use every unique level in group_col with at least 2 samples.

  • method (Literal['welch_anova', 'anova', 'kruskal'] (default: 'welch_anova')) –

    • "welch_anova" (default) — Alexander-Govern test (scipy.stats.alexandergovern), Welch’s generalisation for unequal variances. Robust and recommended.

    • "anova" — classic one-way f_oneway. Assumes equal variances across groups; most sensitive when that holds.

    • "kruskal" — non-parametric kruskal. Use when the Gaussian / symmetry assumption fails even after log.

  • layer (Optional[str] (default: None)) – AnnData layer (default None → adata.X).

Returns:

Indexed by metabolite with columns:

  • stat — test statistic (F, Kruskal H, or Alexander-Govern A)

  • pvalue, padj — raw and BH-FDR

  • mean_<level> — one column per tested group level

  • n_groups — number of levels actually tested

Return type:

pd.DataFrame

Contents