What problem does it solve? Statistical results headed for a paper often ship with incomplete reporting — a p-value without its test statistic, a Markov null without its order k, or prose claims stronger than the numbers support. This Skill provides a structured review pass that catches these gaps after the numbers exist and before prose depends on them. ## Core Features & Use Cases - Reporting-completeness checks: Verifies that test statistics and p-values are reported together, Markov order k is stated, both W2 and persistence landscape L2 appear in diagram-comparison claims, and uncertainty intervals are present. - Lane-audit orchestration: Routes design questions (denominator, exchangeability, clustering, FDR, estimands) to the owning audits such as statistical-design-audit and null-operation-invariance-audit rather than redoing them. - Claim proportionality and provenance: Checks that paper prose is no stronger than the statistics support, that sample counts cite sample_provenance.fitted by stage, and that PROVISIONAL flags are carried through. - Use Case: Before finalizing a P01-A results section, run this review to confirm every permutation test reports both statistic and p-value, every Markov null names its order k, and no provisional result is presented as final. ## Quick Start Review the statistical claims in my P01-B results section and flag any reporting-completeness or proportionality problems before submission.
