Comparison of Semiparametric, Parametric, and Nonparametric ROC Analysis for Continuous Diagnostic Tests Using a Simulation Study and Acute Coronary Syndrome Data

Table 2

Means of the parameter estimates and AUC’s with their standard errors, RMSE, and 95% confidence intervals (CIs) from the parametric, semiparametric, and nonparametric ROC methods using various sample sizes from 1000 simulated datasets generated from the lognormal distribution.

		Parameters
		𝑎=1.400,𝑏=0.900 and 𝐴𝑈𝐶=0.850
n 1 : n 0	Methods	𝑎	𝑆𝐸(𝑎)	𝑏̂	𝑆𝐸(𝑏̂)	𝐴𝑈𝐶	BIAS for 𝐴𝑈𝐶	SE   𝐴𝑈𝐶	RMSE for 𝐴𝑈𝐶	95% CI for 𝐴𝑈𝐶
	P	0.729	0.303	0.236	0.061	0.754	0.096	0.088	0.113	0.581–0.928
15 : 15	S	1.600	0.511	1.176	0.304	0.822	0.028	0.224	0.079	0.383–1.261
	N	—	—	—	—	0.848	0.002	0.070	0.073	0.712–0.986
	P	0.693	0.230	0.216	0.043	0.746	0.104	0.067	0.116	0.610–0.882
25 : 25	S	1.461	0.358	1.020	0.204	0.835	0.015	0.109	0.059	0.622–1.048
	N	—	—	—	—	0.851	0.001	0.054	0.056	0.745–0.957
	P	0.646	0.159	0.196	0.028	0.734	0.116	0.050	0.127	0.635–0.832
50 : 50	S	1.420	0.243	0.955	0.135	0.843	0.007	0.058	0.038	0.728–0.957
	N	—	—	—	—	0.851	0.001	0.038	0.038	0.776–0.926
	P	0.601	0.111	0.183	0.018	0.721	0.129	0.036	0.133	0.650–0.792
100 : 100	S	1.406	0.169	0.930	0.093	0.846	0.004	0.037	0.026	0.773–0.920
	N	—	—	—	—	0.851	0.001	0.027	0.026	0.797–0.904

P: parametric, S: semiparametric, N: nonparametric.