A Two-Stage Penalized Logistic Regression Approach to Case-Control Genome-Wide Association Studies

Table 6

Comparison of TPLR approach and LASSO-patternsearch (the PDR and FDR with subscript M, I, and S indicate the rates calculated for main-effect features, interaction features, and SNPs resp.)

TPLR Approach
𝛾	PDR𝑀	FDR𝑀	PDR𝐼	FDR𝐼	PDR𝑆	FDR𝑆
0.0	0.964	0.902	0.884	0.907	0.947	0.855
0.2	0.768	0.926	0.884	0.900	0.947	0.852
0.4	0.714	0.505	0.748	0.677	0.803	0.494
0.6	0.692	0.199	0.680	0.413	0.730	0.200
0.8	0.684	0.112	0.642	0.331	0.691	0.126
1.0	0.672	0.023	0.610	0.267	0.654	0.073
1.2	0.668	0.021	0.594	0.237	0.638	0.057
1.4	0.658	0.009	0.566	0.201	0.608	0.041
1.6	0.654	0.006	0.536	0.165	0.578	0.026
1.8	0.646	0.006	0.524	0.144	0.563	0.021
2.0	0.630	0.006	0.492	0.109	0.526	0.015
LASSO-patternsearch
𝛼	PDR𝑀	FDR𝑀	PDR𝐼	FDR𝐼	PDR𝑆	FDR𝑆
10^−0	0.882	0.445	0.710	0.967	0.847	0.940
10^−1	0.816	0.332	0.696	0.957	0.827	0.926
10^−2	0.786	0.283	0.664	0.929	0.774	0.885
10^−3	0.718	0.241	0.618	0.869	0.694	0.802
10^−4	0.646	0.220	0.556	0.752	0.609	0.654
10^−5	0.578	0.193	0.486	0.563	0.531	0.444
10^−6	0.504	0.184	0.414	0.355	0.453	0.254
10^−7	0.400	0.190	0.360	0.196	0.383	0.130
10^−8	0.332	0.202	0.292	0.076	0.316	0.047