Table III.

Simulation results for parameter β in univariate model with binary X.

	Settings		Analyses
Model	β	n	PERFECT	CC	NO-T	LOGT	T	T2	NA	COX	COX*
Bias
Base case*	0	336	0.00	−0.01	−0.01	−0.01	−0.01	−0.01	−0.01	−0.01	−0.01
	0.5	336	0.00	0.00	−0.26	−0.06	0.00	−0.02	0.00	0.00	0.00
	1	84	0.01	0.03	−0.52	−0.08	0.03	0.02	0.03	0.04	0.04
Shape 2	1	84	0.01	0.03	−0.52	−0.08	0.00	0.03	0.03	0.04	0.04
Admin cens†	1	84	0.01	0.03	−0.51	0.01	0.02	0.02	0.02	0.02	0.02
Per cent error in model standard error
Base case	0	336	0	1	71	4	−1	0	−1	0	1
	0.5	336	−1	1	73	5	0	−1	0	0	0
	1	84	−2	−6	74	4	−2	−4	−2	−4	−2
Shape 2	1	84	−2	−6	74	4	0	−2	−2	−4	−2
Admin cens	1	84	−4	−1	74	−1	−2	−2	−2	0	0
Power
Base case	0.5	336	92	65	14	59	65	62	65	64	64
	1	84	92	64	12	57	64	60	63	64	65
Shape 2	1	84	92	64	12	57	63	64	63	64	64
Admin cens	1	84	96	73	17	71	73	73	72	71	71

^*Base case: π_M = 0.5, π_X = 0.5, λ_T =0.002, shape κ=1, random censoring with λ_C = 0.002.

^†Administrative censoring.

Monte Carlo error: ≤0.016 for bias, ≤4 per cent for per cent error in model standard error, ≤1.6 per cent for power. Bold cells have bias greater than 0.03, more than 10 per cent error in model standard error or power more than 3 per cent worse than method T.