1[2 {"id": "C1", "text": "Students in treated schools score on average about 0.96 points higher than students in control schools (OLS coefficient).", "results_keys": ["tau_hat", "n_analysed"], "status": "supported", "caveat": "Descriptive difference; causal only if treatment was randomly assigned across schools."},3 {"id": "C2", "text": "At the 5% level, with school-clustered standard errors, the data do not support a non-zero programme effect (p ≈ 0.195; 95% CI [-0.51, 2.44]).", "results_keys": ["se_cluster", "p_value", "tau_ci_low", "tau_ci_high", "effect_significant_5pct"], "status": "supported", "caveat": "Failure to reject is not evidence of zero effect; CR1 with t(39) inference with 40 clusters."},4 {"id": "C3", "text": "The naive i.i.d. standard error understates uncertainty by a factor of about 2.6 because of within-school correlation (residual ICC ≈ 0.235, 25 students per school).", "results_keys": ["se_iid", "se_cluster", "icc_hat"], "status": "supported", "caveat": "ICC is an ANOVA moment estimate with sampling error of its own."}5]6