| Clean re-run (deleted outputs, re-ran analysis.py) gives byte-identical results.json (sha256) |
PASS |
| n_analysed (400) = n_input_rows (400) − exclusions (0) |
PASS |
| se_iid > 0, se_hac > 0 |
PASS |
| se_hac > se_iid (expected with positively autocorrelated x and u) |
PASS (0.0654 vs 0.0448) |
| t_hac = beta_hat / se_hac |
PASS |
| CI contains beta_hat |
PASS |
| p_hac > 0.05 consistent with CI covering 0 |
PASS |
| Memo numbers reproduced (2.0925, 0.1017, 0.0448, 2.27, 0.0238) by our naive baseline |
PASS |
| Bandwidth sensitivity (lags 1–24): p > 0.05 for every lag ≥ 1 |
PASS (p range 0.079–0.125; only lags=0, i.e. White HC, gives p=0.038) |
| rho_resid ≈ 0.61, consistent in sign/magnitude with stated DGP ρ≈0.7 (finite-sample downward bias) |
PASS |