# Verification checks (P4)
1. Clean re-run (deleted results.json, sensitivity.csv, exclusions.csv; re-ran) -> results.json byte-identical (cmp). PASS
2. Exclusions: 400 raw - 0 removed = 400 analysed. PASS
3. se_iid>0, se_hac>0. PASS
4. CI [-0.0261, 0.2296] contains beta_hat. PASS
5. CI contains 0 and p_hac=0.119>0.05 — mutually consistent. PASS
6. t_hac = beta_hat/se_hac exactly. PASS
7. Memo reproduced: alpha 2.0925, beta 0.1017, se 0.0448, t 2.27, p 0.0238 (t ref, df=398). PASS
8. Magnitude: se_hac/se_iid = 1.455 > 1, as expected with positively autocorrelated residuals (rho 0.61) and persistent x. PASS
9. Bandwidth sensitivity: p_normal in [0.096, 0.125] for all lags 5–36; only lags 0–1 give p<0.05. Verdict robust. PASS