| Clean re-run (deleted results.json & exclusions.csv) reproduces byte-identical results.json |
PASS |
| CI [tau_ci_low, tau_ci_high] contains tau_hat |
PASS |
| se_iid > 0, se_cluster > 0 |
PASS |
| n_input_rows − Σ n_removed (exclusions.csv) = n_analysed (1000−0=1000) |
PASS |
| tau_hat = mean_treated − mean_control |
PASS |
| effect_significant_5pct consistent with p_value<0.05 and CI excluding 0 |
PASS |
| Magnitude: se_cluster/se_iid = 2.58 ≈ sqrt(design effect 1+(m−1)ICC)=2.60 |
PASS |
| icc_hat in [0,1] |
PASS |
| Sensitivity: school-means t-test (df 38) gives SE 0.738, p 0.201 — same decision |
PASS |