ggpicrust2

Documentation and diagnostics

• Fixed the existing CRAN check findings: restored generated usage sections for three plotting/comparison helpers and made backend-dependent tests respect unavailable optional packages, including ALDEx2 on macOS oldrel. Help examples also check for their optional analysis/plotting dependencies.
• Restored the installed-package test entry point so R CMD check actually executes the bundled regression tests.
• Camera tutorial examples now estimate within-set correlation explicitly. A fixed correlation of 0.01 can underestimate variance for strongly correlated features; changing the abundance transformation alone does not address this assumption. The public function default is unchanged.
• Restored the documentation website build and deployment from maintained sources, with generated HTML kept out of the source branch.
• GSEA/DAA scatter plots now use the method-specific score label instead of labeling camera/fry signed p-value scores as normalized enrichment scores.
• Audited all help examples, README workflows, and both vignettes against executable examples and underlying data transformations. Examples now use available metadata columns and self-contained inputs; README points to canonical walkthroughs instead of duplicating drifting workflows.
• Clarified method-specific GSEA scores, contrasts, null hypotheses, leading-edge plots, shared tested universes, and the descriptive transformations in PCA, heatmaps, and ridge plots. Annotation adds labels and does not aggregate abundance or convert KO-level p-values to pathway-level evidence.
• Corrected the contribution tutorial: KEGG filters can select member KOs, but the returned contribution table remains KO-level. Added a synthetic example with consistent taxonomy and sample IDs.
• Fixed the stepwise tutorial's pathway_errorbar() call (#207): supply group labels named by sample ID, set KEGG pathway annotation parameters, and skip significance plots when no pathways pass the threshold. Both main workflows now use LinDA; the ALDEx2 alternative selects one test.
• Consolidated duplicated README workflows into the main vignette, corrected unsafe group-vector examples in plotting/table help, and removed FAQ code that rounded p-values or indexed past the available significant features.
• Corrected the GSEA tutorial's KO-versus-pathway comparison and clarified identifier matching for optional taxa-contribution filtering.
• Group-length errors now report the label/sample counts and explain how to supply sample-ID-named labels instead of a metadata column name.