Table 3.

Simulation results for average variance of rejected hypotheses

π0	ρ	m	n	Proposed	Gilbert	Tarone
0.1	0.2	10	100	0.49	0.67	0.72
	0.2	20	100	0.44	0.44	0.46
	0.2	10	1000	1.16	1.95	2.02
	0.2	20	1000	0.51	0.63	0.65
	0.2	10	10000	1.5	4.86	4.97
	0.2	20	10000	0.80	1.79	1.99
	0.5	10	100	0.48	0.86	0.87
	0.5	20	100	0.45	0.69	0.74
	0.5	10	1000	1.35	4.69	6.00
	0.5	20	1000	2.02	8.42	9.47
	0.5	10	10000	5.68	53	60.5
	0.5	20	10000	3.39	54.0	62.58
0.3	0.2	10	100	0.50	0.57	0.60
	0.2	20	100	0.33	0.37	0.39
	0.2	10	1000	0.80	1.27	1.32
	0.2	20	1000	0.55	0.92	0.92
	0.2	10	10000	1.62	4.54	3.70
	0.2	20	10000	0.86	1.65	1.69
	0.5	10	100	0.73	1.8	4.1
	0.5	20	100	0.73	1.46	1.51
	0.5	10	1000	4.3	12.95	13.89
	0.5	20	1000	2.03	11.71	15.09
	0.5	10	10000	4.98	88.8	103.76
	0.5	20	10000	9.5	100.2	146.73

Note: π₀ proportion of true null hypotheses. The number of hypotheses being tested is n, while there are m subjects per group. Proposed represents the procedure outlined in Section 4.1; Gilbert represents the method of Gilbert (2005); Tarone represents the method of Tarone (1990).