Table 2.

Comparison of prediction performances on new data using the selected features

Scenario	ROC AUC				Average accuracy
	SW	EN	LL	BOSS	SW	EN	LL	BOSS
A90	0.86	0.92	0.93	0.96	0.60	0.66	0.71	0.85
A70	0.78	0.82*	0.82*	0.82	0.59	0.60	0.61	0.70
A50	0.64	0.65	0.66	0.67	0.54	0.53	0.53	0.57
B90	0.87	0.92	0.93	0.94	0.60	0.68	0.71	0.80
B70	0.70	0.77*	0.78*	0.77	0.54	0.57	0.58	0.64
B50	0.64	0.68*	0.67*	0.68	0.54	0.54	0.54	0.56

Except where indicated (*), BOSS improved on the three reference methods (p<0.05, one-sided paired t test).

EN, elastic net; LL, logistic lasso; ROC AUC, area under the receiver operating characteristic; SW, stepwise regression.