# Verification checks (P4)

• Clean re-run of out/analysis.py (results/table/figure deleted first): results.json byte-identical — PASS.
• n_train + n_test = n_analysed = 5217 − 215 − 1 = 5001 — PASS; exclusion ledger 5217 − 216 = 5001 — PASS.
• table1 n sums to 5001; all se_bp > 0 — PASS.
• CI [-0.319, 0.420] contains estimate 0.050 — PASS.
• Cross-check of Monday contrast with OLS on Monday dummy, HC1 SE: 0.0502, CI [-0.319, 0.419] — agrees with Welch — PASS.
• Chronology: weekday means used for forecasting estimated on dates ≤ 2009-12-31 only; test dates ≥ 2010-01-01 — PASS (by construction in code).
• Magnitudes: RMSE ~4.9 bp/day plausible for 10y yield; model RMSE ≥ baseline by 0.008 bp — consistent with null ANOVA.