| Clean re-run (deleted results.json, exclusions.csv; reran) gives byte-identical results.json |
PASS |
| tau_hat equals treated-minus-control difference of student means |
PASS |
| se_iid > 0, se_cluster > 0 |
PASS |
| CI [tau_ci_low, tau_ci_high] contains tau_hat |
PASS |
| effect_significant_5pct == (p_value < 0.05) |
PASS (false; CI also contains 0) |
| icc_hat in [0,1] |
PASS (0.240) |
| n_analysed = n_raw − removed = 1000; 40×25 = 1000 |
PASS |
| se_cluster/se_iid (2.584) ≈ sqrt Moulton design effect (2.601) |
PASS (magnitude consistent with ICC) |
| Cluster SE vs school-means two-sample t-test: SE 0.729 vs 0.738, p 0.195 vs 0.201 |
PASS (same conclusion) |