Table 1.

Average empirical Type I error at nominal α = 0.05. Italicized values indicate that the empirical Type I error is greater than 0.055.

METHOD	β= 0	β= 0.5	β= 1	β= 2
n1 = 70, n2 = 15, n3 = 15
Gold-std	0.0455	0.0473	0.0503	0.0501
Paired only	0.0479	0.0494	0.0490	0.0518
Two-sample	0.0475	0.0643	0.0794	0.0922
FM-PS	0.0462	0.0485	0.0469	0.0476
Reg-PS	0.0480	0.0502	0.0441	0.0442
n1 = 50, n2 = 25, n3 = 25
Gold-std	0.0527	0.0541	0.0495	0.0541
Paired only	0.0486	0.0536	0.0534	0.0531
Two-sample	0.0476	0.1044	0.1460	0.1809
FM-PS	0.0483	0.0467	0.0465	0.0485
Reg-PS	0.0494	0.0476	0.0443	0.0416
n1 = 30, n2 = 35, n3 = 35
Gold-std	0.0522	0.0543	0.0483	0.0489
Paired only	0.0507	0.0494	0.0499	0.0465
Two-sample	0.0522	0.1712	0.2805	0.3663
FM-PS	0.0449	0.0454	0.0422	0.0460
Reg-PS	0.0524	0.0479	0.0452	0.0446