| name | stats-sanity |
| description | Audit statistical consistency in manuscripts, reports, experiments, and result tables. Use for Stats Sanity, p-value checks, t/F/chi-square/r/z consistency, GRIM/GRIMMER/DEBIT-style checks, denominator consistency, effect sizes, confidence intervals, multiple comparisons, or graph/table numeric consistency. |
Stats Sanity
Goal: check whether reported statistics are internally consistent and appropriate for the design.
Checklist:
- Extract all reported metrics, denominators, sample sizes, means, SD/SE, confidence intervals, test statistics, p-values, and correction methods.
- Recompute simple statistics when enough information is available.
- Check p-values against test statistics and degrees of freedom.
- Check table/figure/text consistency and denominator drift.
- Flag multiple-comparison, one-tailed/two-tailed, effect-size, and confidence-interval omissions.
- Report issues as
blocking, likely error, needs clarification, or style/reporting.
Do not use LLM mental math for final numeric checks; use code or a calculator when numbers matter.
