# Verification checks (P4)

• Clean re-run of out/analysis.py → byte-identical out/results.json: PASS
• CI [beta_ci_low, beta_ci_high] contains beta_hat: PASS
• se_iid > 0 and se_hac > 0: PASS
• n_analysed (400) = n_input_rows (400) − removed (0): PASS
• Memo numbers reproduced (intercept 2.0925, slope 0.1017, SE 0.0448, t 2.27, p 0.0238): PASS
• se_hac > se_iid, consistent with positively autocorrelated residuals and persistent x: PASS
• t_hac = beta_hat / se_hac: PASS
• CI contains 0 ⇔ p_hac ≥ 0.05: PASS
• Verdict (p_hac > 0.05) robust across bandwidths 5, 6, 8, 9, 12, 18, 24: PASS