Table 1.

Simulation results evaluating the effect of the proportion of samples used in stage 1, P, and the threshold for continuing to stage 2, ROC_co (0.1) on the operating characteristics of our study. Simulations were completed using n_s = 5, W_s = (1, …, 1), V_s = (1, …, 1) and ρ = 0.0. 10000 simulations were completed for each scenario.

	Null hypothesis1					Alternative hypothesis2
P	% Early Stop	E(SS)	Cond α	Uncond α	ROC (0.1) for Xtβ̂1	% Early Stop	E(SS)	Cond 1 – β	Uncond 1 – β	ROC (0.1) for Xtβ̂1
Fixed-sample
0.3	0.00	720	0.004	0.004	0.371	0.00	720	0.905	0.905	0.628
0.4	0.00	720	0.007	0.007	0.379	0.00	720	0.888	0.888	0.635
0.5	0.00	720	0.011	0.011	0.384	0.00	720	0.838	0.838	0.638
0.6	0.00	720	0.012	0.012	0.387	0.00	720	0.754	0.754	0.640
0.7	0.00	720	0.012	0.012	0.389	0.00	720	0.646	0.646	0.642
ROCco(0.1) = 0.40
0.3	0.43	502	0.006	0.003	0.372	0.00	718	0.908	0.904	0.629
0.4	0.46	521	0.007	0.004	0.380	0.00	719	0.883	0.882	0.635
0.5	0.48	546	0.011	0.006	0.385	0.00	720	0.832	0.832	0.638
0.6	0.50	576	0.011	0.006	0.387	0.00	720	0.756	0.756	0.640
0.7	0.50	612	0.016	0.008	0.389	0.00	720	0.635	0.635	0.642
ROCco(0.1) = 0.45
0.3	0.64	397	0.004	0.002	0.372	0.02	712	0.909	0.894	0.628
0.4	0.69	423	0.010	0.003	0.380	0.01	717	0.887	0.880	0.635
0.5	0.72	459	0.010	0.003	0.385	0.00	719	0.834	0.830	0.638
0.6	0.76	502	0.015	0.004	0.387	0.00	719	0.754	0.752	0.640
0.7	0.79	550	0.017	0.004	0.389	0.00	720	0.637	0.637	0.642
ROCco(0.1) = 0.50
0.3	0.81	311	0.005	0.001	0.373	0.05	697	0.904	0.863	0.629
0.4	0.85	352	0.009	0.001	0.380	0.03	706	0.879	0.851	0.635
0.5	0.89	398	0.009	0.001	0.384	0.02	711	0.837	0.817	0.638
0.6	0.92	454	0.008	0.001	0.388	0.01	716	0.753	0.743	0.640
0.7	0.94	516	0.013	0.001	0.389	0.01	718	0.634	0.627	0.642

¹ROC (0.1)₀ = 0.40

²ROC (0.1)_a = 0.65