1# Verification checks (P4)2| check | outcome |3|---|---|4| Clean re-run (results.json, exclusions.csv deleted) regenerates byte-identical results.json | PASS |5| CI [tau_ci_low, tau_ci_high] contains tau_hat | PASS |6| se_iid > 0, se_cluster > 0 | PASS |7| tau_hat equals mean_score_treated − mean_score_control | PASS |8| effect_significant_5pct consistent with p_value<0.05 and CI excluding 0 | PASS (all false) |9| n_input_rows 1000 − 0 removed = n_analysed 1000 (exclusions.csv sums) | PASS |10| icc_hat in [0,1] | PASS (0.240) |11| (se_cluster/se_iid)^2 = 6.68 vs design effect 1+(25−1)·icc = 6.77 (within 10%) | PASS |12| se_cluster (CR1) vs Welch SE on 40 school means 0.738 (within 5%); Welch p=0.201 vs cluster p=0.195 | PASS |13| treated constant within school; 20/20 schools; balanced 25 per school | PASS (audit) |14