Table 1.

Pearson correlation between the true and predicted genomic breeding values for group bridge, MCP, lasso and SCAD for trait T1 based on systematic groups.

Group size	Penalty selected by AIC					Penalty selected by BIC
	Bridge	MCP	Lasso	SCAD	SGLasso	Bridge	MCP	Lasso	SCAD
1	0.682	0.758	0.778	0.767	0.781	0.682	0.768	0.773	0.652
10	0.770	0.759	0.793	0.788	0.787	0.753	0.772	0.747	0.735
20	0.774	0.761	0.788	0.770	0.787	0.777	0.772	0.756	0.667
30	0.758	0.760	0.790	0.774	0.787	0.787	0.771	0.753	0.644
40	0.769	0.761	0.789	0.758	0.787	0.771	0.772	0.754	0.595
50	0.774	0.761	0.780	0.740	0.791	0.763	0.770	0.732	0.579
60	0.765	0.760	0.784	0.750	0.791	0.765	0.771	0.706	0.581
70	0.771	0.760	0.779	0.760	0.789	0.757	0.770	0.718	0.619
80	0.781	0.761	0.776	0.759	0.795	0.747	0.770	0.721	0.550
90	0.761	0.760	0.774	0.748	0.781	0.743	0.770	0.709	0.478
100	0.771	0.760	0.778	0.706	0.790	0.760	0.770	0.704	0.502

Although listed under AIC, SGLasso used only 10-fold cross validation. The Pearson correlation for ridge regression for comparison is 0.737.