Table 3.

Power for the unadjusted and adjusted tests under simple versus constrained randomization using imbalance score (B) with g = 7 and g = 11. All four potential group-level confounders were adjusted in constrained randomization (S = 4) and in any given adjusted test; candidate set size (R) are varied under constrained randomization.

				Power
	Randomization	ICC	R	Unadj F-test	Unadj P-test	Adj F-test	Adj P-test
g = 7	Constrained	0.01	20	0.013	–	1.000	–
	Constrained	0.01	100	0.015	0.579	1.000	0.999
	Constrained	0.01	1000	0.040	0.270	1.000	0.999
	Constrained	0.01	2000	0.086	0.236	1.000	0.996
	Constrained	0.01	3000	0.123	0.170	0.999	0.980
	Simple	0.01	–	0.148	0.143	0.996	0.946
	Constrained	0.1	20	0.022	–	0.638	–
	Constrained	0.1	100	0.025	0.371	0.631	0.544
	Constrained	0.1	1000	0.046	0.209	0.608	0.576
	Constrained	0.1	2000	0.084	0.196	0.574	0.536
	Constrained	0.1	3000	0.118	0.158	0.546	0.511
	Simple	0.1	–	0.137	0.135	0.516	0.477
g = 11	Constrained	0.01	20	0.007	–	1.000	–
	Constrained	0.01	100	0.010	0.678	1.000	1.000
	Constrained	0.01	1000	0.026	0.719	1.000	1.000
	Constrained	0.01	5000	0.105	0.519	1.000	1.000
	Constrained	0.01	10000	0.153	0.393	1.000	1.000
	Simple	0.01	–	0.223	0.220	1.000	0.999
	Constrained	0.1	20	0.022	–	0.898	–
	Constrained	0.1	100	0.024	0.464	0.896	0.839
	Constrained	0.1	1000	0.041	0.526	0.888	0.882
	Constrained	0.1	5000	0.106	0.413	0.883	0.876
	Constrained	0.1	10000	0.145	0.331	0.875	0.865
	Simple	0.1	–	0.202	0.201	0.827	0.800