Table 3.

A comparison of ARMS with AIC, full, and best univariate models in logistic regression for prostate cancer data (n = 200) under 1000 random permutations

Method	L2 risk	L1 risk	EP	AUC
ARMS-LIKELI	0.083	0.169	0.388	0.79
	(0.002)	(0.004)	(0.006)	(0.01)
ARMS-AIC	0.082	0.166	0.382	0.80
	(0.002)	(0.004)	(0.006)	(0.01)
ARMS-GDF	0.078	0.159	0.373	0.82
	(0.002)	(0.004)	(0.006)	(0.01)
ARMS-APE	0.087	0.175	0.398	0.77
	(0.003)	(0.005)	(0.007)	(0.01)
ARMS-RESID	0.088	0.177	0.401	0.76
	(0.003)	(0.005)	(0.007)	(0.01)
AIC	0.089	0.179	0.405	0.76
	(0.003)	(0.005)	(0.007)	(0.01)
Full	0.095	0.193	0.418	0.73
	(0.003)	(0.005)	(0.007)	(0.01)
Univariate (AMACR)	0.106	0.214	0.439	0.70
	(0.003)	(0.006)	(0.008)	(0.01)