Novel Harmonic Regularization Approach for Variable Selection in Cox’s Proportional Hazards Model

Table 1

Average number of the variable selected and the recovery rate by the seven regularization methods on the simulated data in 500 runs.

Corr.	Size	Average of variable selected
			SCAD	MCP	A-Lasso	Len		HRA
	100	33.4	11.6	9.2	17.8	35.8	6.6	6.6
	150	29.7	9.8	7.9	12.8	31.2	5.9	6.3
	200	24.4	8.4	6.1	9.4	24.8	5.7	5.8
	100	43.2	14.8	11.7	22.9	46.9	9.9	9.7
	150	34.5	11.2	8.7	16.5	36.3	7.2	7.3
	200	26.7	9.7	7.7	11.4	28.2	6.5	7
	100	45.1	15.1	12.2	27.2	47.8	10.8	10.9
	150	39.3	11.9	10.8	20.1	44.3	8.4	8.3
	200	27.1	10.1	8.4	12.6	30.6	7.3	7.7
	100	55.6	17.7	16	32.7	56.3	13.9	15.2
	150	47.3	15.9	14.8	25.9	48.6	9.4	10
	200	36.8	13.7	12.5	19.5	41.6	7.8	7.8
Corr.	Size	Recovery rate
			SCAD	MCP	A-Lasso	Len		HRA
	100	0.14	0.43	0.54	0.28	0.13	0.71	0.72
	150	0.16	0.51	0.63	0.39	0.16	0.8	0.79
	200	0.2	0.59	0.81	0.53	0.2	0.87	0.86
	100	0.11	0.33	0.42	0.21	0.1	0.5	0.51
	150	0.14	0.44	0.57	0.3	0.13	0.65	0.63
	200	0.18	0.51	0.64	0.43	0.17	0.7	0.71
	100	0.11	0.33	0.4	0.18	0.1	0.46	0.45
	150	0.12	0.42	0.46	0.24	0.11	0.59	0.6
	200	0.18	0.49	0.59	0.39	0.16	0.62	0.63
	100	0.08	0.28	0.31	0.15	0.08	0.35	0.32
	150	0.1	0.31	0.33	0.19	0.1	0.51	0.5
	200	0.13	0.36	0.4	0.25	0.12	0.61	0.61