← graph

validate-statistics-fdr-cis

Validate: statistics audit (BH-FDR, CIs, effect sizes)

Metadata

Statusdone
Assignedagent-695
Agent identity3184716484e6f0ea08bb13539daf07686ee79d440505f1fdf2de0357707034c3
Created2026-05-05T05:08:32.804072409+00:00
Started2026-05-05T05:28:08.920907643+00:00
Completed2026-05-05T05:44:41.202176013+00:00
Tagspaper-prep,validation, eval-scheduled
Eval score0.90
└ blocking impact0.88
└ completeness0.94
└ constraint fidelity0.85
└ coordination overhead0.85
└ correctness0.94
└ downstream usability0.90
└ efficiency0.86
└ intent fidelity0.96
└ style adherence0.78

Description

Description

Statistics audit on all main-text claims in paper_prep/synthesis/MANUSCRIPT_SKELETON.md. Implement directly — do not decompose further.

Validation

  • Every p-value in MANUSCRIPT_SKELETON.md headline-numbers section has either a BH-corrected q-value column or an explicit "uncorrected" annotation
  • Every odds-ratio (e.g. f7501 chr16_q OR = 17.4) has a 95% CI (Fisher mid-p exact or bootstrap)
  • Multi-resolution Mantel ρ aggregated to a single SI table (5/10/20/50/100 kb × 8 datasets)
  • f7501 per-arm × per-superpopulation Fisher tests (~65) corrected with BH-FDR (SURVEY_01 §5 #1)
  • Output: paper_prep/synthesis/STATS_AUDIT.md; per-test corrections written into source TSVs alongside

Inputs

  • paper_prep/synthesis/MANUSCRIPT_SKELETON.md, WORK_DECOMPOSITION.md
  • All 13 paper_prep/surveys/SURVEY_*.md

Depends on

Required by

Log