Table 1.

Type I Error (%), Scenario 1: Uncensored data, 4 outcomes. n, m = sample size per strata in each group, respectively. T_O = O'Brien original test; $T_O^U=\mathit{\text{Proposed O'Brien test}}$; $T_O^{U,\mathit{\text{Ad}}}=\mathit{\text{Proposed Adaptive O'Brien test}}$. ρ = correlation between outcomes; for unequal variances below, group 1 covariance Σ₁ = diag(1, 1, 1, 1); for group 2, Σ₂ has elements (1,9,16,25) on the diagonal, and Σ_2ij = 1 for i ≠ j.

Variances	ρ	No. Strata	n	m	TO	$T_O^U$	$T_O^{U,\mathit{\text{Ad}}}$
Equal	0	2	15	15	4.7	4.2	4.3
			30	30	5.4	5.0	5.8
			100	100	4.9	4.8	5.0
			80	40	5.0	4.8	5.1
		4	15	15	5.5	5.7	6.0
			30	30	5.1	5.1	5.4
			100	100	5.0	5.0	5.1
			80	40	4.3	4.2	5.1
	0.5	2	15	15	5.6	5.0	4.9
			30	30	5.2	4.7	4.9
			100	100	5.4	5.3	5.4
			80	40	5.1	4.9	5.0
		4	15	15	5.6	5.5	5.5
			30	30	4.9	5.0	5.3
			100	100	5.3	5.3	5.6
			80	40	5.7	5.7	5.4
Unequal	See Caption	2	15	15	6.5	4.6	4.7
			30	30	6.5	4.8	4.9
			100	100	6.6	4.9	5.3
			80	40	6.0	4.7	5.3
		4	15	15	7.2	5.5	5.9
			30	30	6.7	5.0	5.2
			100	100	6.6	4.8	5.2
			80	40	5.9	4.8	4.9