Table 4.

Performance of the nonparametric ROC(t) and ROC(t) estimators. Standard errors of biases of ROC(t) or AROC(t) are less than 0.001. Standard errors for coverage of 95% confidence intervals do not exceed 0.7% of the value. Standard errors for relative efficiency do not exceed 1% of the value.

	Population 1	Population 2	Combined-data
Bias × 1000
t = 0.1	0.012	0.014	0.008
t = 0.3	0.005	0.008	0.002
t = 0.5	0.004	0.006	0.002
t = 0.7	0.004	0.006	0.002
t = 0.9	0.004	0.006	0.002
Coverage of 95% bootstrap percentile CI
t = 0.1	94.4	95.4	95.6
t = 0.3	95.4	94.6	95.7
t = 0.5	94.8	94.2	95.6
t = 0.7	94.0	92.4	95.3
t = 0.9	90.2	82.9	93.8
Efficiency of CML relative to the nonparametric estimator
t = 0.1	1.83	1.64	1.84
t = 0.3	1.21	1.23	1.26
t = 0.5	1.28	1.29	1.32
t = 0.7	1.22	1.23	1.25
t = 0.9	1.72	1.68	1.88