Skip to main content
. Author manuscript; available in PMC: 2021 Aug 30.
Published in final edited form as: Stat Med. 2020 May 3;39(19):2568–2586. doi: 10.1002/sim.8557

TABLE 5.

Coverage probability and type I error rate (Weibull distribution, γ = 0.5)

Selected Coverage (Type I error rate)a
True log hazard ratios partitions (S) Naive Duality
All 94.8 (4.5) 98.3 (1.7)
θ1 = θ2 = θ3 = θ4 =0.0198 1, 2, and 3 96.1 (3.0) 99.1 (0.9)
1 and 2 96.0 (3.1) 99.2 (0.8)
1 96.3 (2.9) 99.6 (0.4)
All 93.0 (6.6) 98.4 (1.5)
θ1 = −0.2231, θ2 = −0.0953, 1, 2, and 3 96.0 (3.1) 99.1 (0.5)
θ3 = 0.3365, θ4 = 0.4055 1 and 2 96.3 (1.4) 99.2 (0.0)
1 96.1 (1.5) 99.4 (0.0)
All 95.4 (2.6) 98.7 (0.7)
θ1 = −0.4055, θ2 = −0.2231, 1, 2, and 3 93.3 (0.3) 97.6 (0.0)
θ3 = −0.0953, θ4 = 0 1 and 2 93.8 (0.7) 98.2 (0.0)
1 94.0 (0.9) 98.6 (0.0)
a

Type I error is the probability that at least one upper bound is less than the true value.