# Verification checks (P4)
| check | outcome |
|---|---|
| Clean re-run (results.json, exclusions.csv deleted) regenerates byte-identical results.json | PASS |
| CI [tau_ci_low, tau_ci_high] contains tau_hat | PASS |
| se_iid > 0, se_cluster > 0 | PASS |
| tau_hat equals mean_score_treated − mean_score_control | PASS |
| effect_significant_5pct consistent with p_value<0.05 and CI excluding 0 | PASS (all false) |
| n_input_rows 1000 − 0 removed = n_analysed 1000 (exclusions.csv sums) | PASS |
| icc_hat in [0,1] | PASS (0.240) |
| (se_cluster/se_iid)^2 = 6.68 vs design effect 1+(25−1)·icc = 6.77 (within 10%) | PASS |
| se_cluster (CR1) vs Welch SE on 40 school means 0.738 (within 5%); Welch p=0.201 vs cluster p=0.195 | PASS |
| treated constant within school; 20/20 schools; balanced 25 per school | PASS (audit) |