Table 3.

Simulation results for comparison of semi-parametric approach of Heagerty and Zheng (2005) (HZ2005) and non-parametric approach WMR. We assumed follows a mixture of two multivariate normal distributions: if Z = 1 and if Z = 0, where Z∼ Bernoulli (0.2). An independent censoring time was generated such that 20% of the subjects were censored. We show the estimated HZ2005, WMR, MCSDs, and the SD estimated using the proposed variance estimator (EstSD) and the coverage (nominal: 95.0) for WMR

Log time	AUC(t)	HZ2005	MCSD	WMR	MCSD	EstSD	Coverage
− 2.5	0.378	0.173	0.069	0.376	0.109	0.097	87.7
− 2.0	0.481	0.346	0.087	0.477	0.083	0.075	91.1
− 1.5	0.591	0.551	0.071	0.595	0.062	0.057	92.0
− 1.0	0.673	0.660	0.048	0.674	0.048	0.043	91.6
− 0.5	0.709	0.689	0.032	0.708	0.035	0.034	93.7
0.0	0.709	0.684	0.023	0.710	0.030	0.030	94.8
0.5	0.691	0.666	0.022	0.692	0.034	0.033	92.5
1.0	0.669	0.646	0.022	0.669	0.042	0.041	93.9