Table 1. Mean results of the simulation.

In bold–the best performance amongst all the methods.

		Scenario
ρ	Method	1	2	3	4	1	2	3	4	1	2	3	4
		Sensitivity of feature selection				Specificity of feature selection				Accuracy of classification (test set)
	Lasso	0.966	0.798	0.344	0.361	0.996	0.968	0.967	0.966	89.26%	81.47%	84.76%	80.26%
	L1/2	0.971	0.888	0.411	0.355	0.998	0.974	0.975	0.970	92.05%	82.22%	85.11%	81.45%
0.3	SCAD − L2	1.000	0.913	0.722	0.674	0.995	0.928	0.890	0.723	93.21%	82.90%	84.51%	82.51%
	EN	0.997	0.916	0.737	0.662	0.994	0.926	0.886	0.735	91.03%	81.34%	84.47%	80.27%
	HLR	1.000	0.924	0.791	0.708	0.999	0.931	0.892	0.769	95.27%	82.66%	84.99%	85.05%
	Lasso	0.887	0.723	0.351	0.270	0.995	0.975	0.981	0.923	94.24%	84.10%	91.88%	85.88%
	L1/2	0.755	0.630	0.275	0.220	1.000	0.974	0.988	0.928	95.90%	86.50%	90.20%	84.20%
0.6	SCAD − L2	1.000	0.866	0.800	0.629	1.000	0.949	0.929	0.849	96.33%	86.43%	89.20%	93.03%
	EN	1.000	0.854	0.795	0.621	1.000	0.953	0.939	0.837	96.22%	86.41%	92.12%	91.01%
	HLR	1.000	0.875	0.816	0.636	1.000	0.968	0.942	0.841	99.53%	87.16%	92.71%	92.82%
	Lasso	0.548	0.548	0.174	0.145	0.938	0.972	0.987	0.934	96.05%	86.79%	93.22%	91.15%
	L1/2	0.337	0.495	0.159	0.139	0.999	0.977	0.991	0.944	97.89%	87.90%	93.70%	92.70%
0.9	SCAD − L2	1.000	0.872	0.809	0.636	1.000	0.954	0.952	0.861	97.28%	88.60%	93.70%	93.19%
	EN	1.000	0.856	0.818	0.622	0.995	0.951	0.949	0.875	98.22%	88.14%	93.52%	93.82%
	HLR	1.000	0.897	0.824	0.645	1.000	0.966	0.956	0.880	99.87%	89.40%	94.76%	94.40%

Mean results are based on 500 repeats. The sensitivity and specificity are both dedicated to measures the quality of the selected features, the accuracy evaluates the classification performance of the different regularization approaches on the test sets.