Table II.

Large sample (a) and small sample (b) performance of the method in estimating the biological efficacy (ψ₁) in a Weibull model under various censoring rates (CR) and treatable subgroup proportions (p), (c) simulation of MSLT-I conditions (CR>75 per cent, p = 0.2) under three different efficacy scenarios.

CR	True p	ψ̂1	Coverage	Simulation variance	Estimated variance
(a) Large sample simulation (N = 1000, true ψ1 = 0.85)
40 per cent	0.3	0.8432	94.7	0.0420	0.0421
	0.5	0.8551	95.1	0.0197	0.0199
	0.8	0.8502	94.7	0.0106	0.0106
60 per cent	0.3	0.8539	95.3	0.0477	0.0487
	0.5	0.8521	95.3	0.0243	0.0247
	0.8	0.8492	95.5	0.0141	0.0144
(b) Small sample simulation (N = 400, true ψ1 = 0.85)
40 per cent	0.3	0.8379	93.9	0.1176	0.1108
	0.5	0.8504	93.8	0.0552	0.0500
	0.8	0.8523	95.9	0.0257	0.0266
60 per cent	0.3	0.8414	93.5	0.1325	0.1255
	0.5	0.8565	93.8	0.0632	0.0645
	0.8	0.8544	94.6	0.0364	0.0362
(c) Simulation of MSLT-I (N = 1300)
True ψ1
1.00		0.9842	94.7	0.1023	0.1012
0.50		0.5012	95.6	0.0950	0.0957
0.00		0.0071	95.7	0.0923	0.0966