1[2 {"id":"C1","text":"CR0 with normal critical value under-covers in every cell and fails (coverage < 0.90) in all G=5 and G=10 cells.","results_keys":["coverage","failure_cells"],"status":"supported","caveat":"Gaussian DGP, one cluster-size pattern."},3 {"id":"C2","text":"All three procedures fail at G=5 except CR1 with ICC=0 (0.920); no procedure fails at G=30.","results_keys":["coverage","failure_cells","n_failure_cells"],"status":"supported","caveat":"CR1 G=5 ICC 0.2/0.5 are only ~2.5-3 MC SE below 0.90."},4 {"id":"C3","text":"CR1 has the highest average coverage over the G=5 cells.","results_keys":["best_procedure_small_G","coverage"],"status":"supported","caveat":"WCB implemented with Rademacher weights and B=399; with G=5 only 32 distinct weight vectors exist."},5 {"id":"C4","text":"CR0 and CR1 variance estimators are biased downward in all cells, capturing only ~50% (CR0) and ~63% (CR1) of Var(ybar) at G=5 with ICC>0, approaching 0.92-0.97 at G=30.","results_keys":["rel_bias_var"],"status":"supported","caveat":"MC SE of relative bias 0.007-0.016."},6 {"id":"C5","text":"Only 5 pairs are consistent with nominal coverage, all at G=30 (CR1 at ICC 0 and 0.2, WCB at all ICC).","results_keys":["nominal_cells","coverage","mc_se_coverage"],"status":"supported","caveat":"Pointwise 2-SE criterion, no multiplicity adjustment; borderline pairs could flip under another seed."},7 {"id":"C6","text":"At G=10 with ICC>0, WCB is closest to nominal (0.937-0.939) versus CR1 (0.917).","results_keys":["coverage"],"status":"supported","caveat":"Differences of ~2 points, roughly 2.5-3 combined MC SE."},8 {"id":"C7","text":"WCB under-coverage at G=5 is due to discreteness of the Rademacher bootstrap distribution and q being small when one large cluster dominates.","results_keys":[],"status":"not_evaluated","caveat":"Plausible mechanism, not tested separately."}9]10