# Verification checks (P4)

1. Clean re-run (deleted results.json, table1.csv, figure1.png, exclusions.csv; re-ran analysis.py): results.json byte-identical, table1.csv identical — PASS.
2. ANOVA recomputed by hand (SSB/SSW): F = 0.263466, p = 0.9015 — matches scipy — PASS.
3. Counts: sum(table n) = n_analysed = n_train + n_test = 5217 − 215 − 1 = 5001 — PASS. Exclusion ledger adds up — PASS.
4. Monday contrast recomputed from table1 weighted means = 0.050170 = effect_estimate — PASS.
5. CI contains estimate; all SEs > 0 — PASS. HC1 regression CI ([-0.319, 0.419]) essentially equal to Welch CI — PASS (robustness).
6. Information rule: train = dates ≤ 2009-12-31, test ≥ 2010-01-01 (asserted, disjoint, exhaustive); forecast means from train only — PASS.
7. Magnitudes: weekday means ~±0.2 bp vs daily sd ~5.7 bp; model RMSE 4.884 vs baseline 4.876 bp — plausible — PASS.
8. Figure inspected visually: five points, 95% bars, all overlapping zero — PASS.