TABLE 2.

Simulation results under g(t) = exp(t) with covariate-dependent censoring.

		α1				α2				$\omega ({\beta }_{*},\widehat{\beta })$
	N	Bias	SD	SE	CP	Bias	SD	SE	CP	Mean	SD
S1. Linear case
IPCW1	2000	−0.096	0.099	0.097	85.8	−0.090	0.092	0.098	84.6	13.53	4.90
IPCW2	2000	0.003	0.090	0.089	93.8	−0.004	0.093	0.089	94.0	7.19	3.22
DR	2000	0.004	0.091	0.091	93.7	0.006	0.090	0.091	94.6	7.36	3.26
QPS	2000	0.004	0.102	0.091	93.8	0.004	0.093	0.091	94.7	7.94	3.56
IPCW1	4000	−0.096	0.073	0.072	72.8	−0.095	0.072	0.073	69.2	12.44	3.45
IPCW2	4000	0.002	0.064	0.063	95.6	0.001	0.065	0.062	93.6	4.78	2.08
DR	4000	0.003	0.066	0.066	95.4	0.004	0.067	0.066	94.1	4.96	2.14
QPS	4000	0.002	0.073	0.074	94.4	0.001	0.070	0.072	96.3	5.38	2.30
S2. Sine curve case
IPCW1	2000	−0.085	0.085	0.087	83.8	−0.081	0.090	0.087	83.4	5.88	2.01
IPCW2	2000	−0.006	0.076	0.079	95.6	−0.002	0.078	0.079	95.4	3.42	1.37
DR	2000	−0.002	0.082	0.083	93.5	0.004	0.083	0.083	94.3	3.37	1.44
QPS	2000	−0.009	0.096	0.099	95.2	−0.001	0.096	0.099	96.2	3.82	1.67
IPCW1	4000	−0.081	0.065	0.067	74.2	−0.081	0.066	0.068	75.3	5.50	1.56
IPCW2	4000	0.001	0.056	0.056	95.8	0.001	0.057	0.056	94.2	2.29	0.97
DR	4000	0.002	0.062	0.062	94.7	0.002	0.062	0.063	93.8	2.31	0.98
QPS	4000	0.003	0.072	0.072	95.6	−0.001	0.069	0.070	94.8	2.54	1.12
S3. Quadratic case
IPCW1	2000	−0.087	0.100	0.097	87.2	−0.082	0.093	0.095	88.3	12.02	5.89
IPCW2	2000	−0.005	0.083	0.087	95.8	0.001	0.087	0.087	94.2	11.62	5.23
DR	2000	0.002	0.090	0.091	96.4	0.004	0.089	0.089	94.7	11.54	5.63
QPS	2000	−0.002	0.096	0.095	93.7	0.002	0.091	0.092	95.6	13.07	7.07
IPCW1	4000	−0.082	0.069	0.069	75.3	−0.081	0.070	0.071	76.2	8.92	3.85
IPCW2	4000	0.004	0.062	0.061	93.4	−0.002	0.062	0.061	94.0	8.13	3.62
DR	4000	0.004	0.064	0.064	94.3	0.005	0.064	0.065	95.5	8.40	3.72
QPS	4000	0.001	0.070	0.070	95.7	0.001	0.067	0.070	94.6	9.21	4.30

SD: sample standard deviation; SE: mean of estimated standard error; CP: empirical coverage probability of 95% confidence interval; IPCW₁: inverse-probability-of-censoring weighting estimator with a mis-specified censoring distribution; IPCW₂: inverse-probability-of-censoring weighting estimator with a correctly specified censoring distribution; DR: double-robust estimator; QPS: quasi-partial score estimator; $\omega \left({\beta }_{*},\widehat{\beta }\right)$ was calculated by arccos $\left(〈{\beta }_{*},\widehat{\beta }〉/\Vert \widehat{\beta }\Vert \cdot ‖{\beta }_{*}‖\right)$; Censoring rate was 10%.