1[2 {"id":"C1","text":"The estimated programme effect is 0.961 points (treated minus control student-weighted mean).","results_keys":["tau_hat","mean_treated","mean_control"],"status":"supported","caveat":"Causal reading assumes as-if random assignment of schools."},3 {"id":"C2","text":"With school-clustered uncertainty, the programme effect is not statistically different from zero at the 5% level (p=0.195, 95% CI [-0.513, 2.436]).","results_keys":["se_cluster","p_value","tau_ci_low","tau_ci_high","effect_significant_5pct"],"status":"supported","caveat":"Non-significance is not evidence of no effect; CI includes effects up to ~2.4 points."},4 {"id":"C3","text":"The naive iid standard error understates uncertainty by a factor of about 2.6, consistent with a residual ICC of 0.24 and 25 students per school.","results_keys":["se_iid","se_cluster","se_ratio_cluster_to_iid","icc_hat","design_effect_from_icc"],"status":"supported","caveat":"ICC is an ANOVA moment estimate."},5 {"id":"C4","text":"A school-means analysis gives the same conclusion (SE 0.738, p 0.201).","results_keys":["school_means_se","school_means_p_value"],"status":"supported","caveat":"Equivalence of point estimates relies on the balanced design."}6]7