Resampling-based tests for Lasso in genome-wide association studies

BMC Genomic Data

Table 1 Comparison of methods given fixed sample size and increasing number of null SNPs, α=0.01

Model	0 Null SNPs	45 Null SNPs	300 Null SNPs	900 Null SNPs	20,000 Null SNPs
Lasso-PL	0.840	0.831 (0.008)	0.841 (0.0081)	0.857 (0.0083)	0.841 (0.0095)
Lasso-AL	0.850	0.850 (0.0102)	0.851 (0.0089)	0.854 (0.0077)	0.828 (0.0079)
Lasso-Ayers	0.819	0.808 (0.0121)	0.836 (0.0096)	0.846 (0.0086)	0.838 (0.0095)
SMA	0.828	0.828 (0.0106)	0.828 (0.0101)	0.828 (0.01)	0.828 (0.01)
Lasso-PT	0.829	0.832 (0.0078)	0.827 (0.0073)	0.818 (0.0076)	0.560 (0.0013)
Lasso-RB	0.869	0.859 (0.0137)	0.847 (0.0106)	0.833 (0.0089)	0.555 (0.0012)
Lasso-MRB(t=0.001)	0.869	0.865 (0.0138)	0.850 (0.0105)	0.838 (0.0089)	0.556 (0.0012)
Lasso-MRB(t=0.005)	0.869	0.863 (0.0137)	0.849 (0.0105)	0.836 (0.009)	0.555 (0.0012)
Lasso-MRB(t=0.01)	0.869	0.864 (0.0139)	0.849 (0.0105)	0.837 (0.009)	0.556 (0.0012)
Lasso-MRB(t=0.03)	0.869	0.864 (0.0136)	0.849 (0.0102)	0.839 (0.0092)	0.558 (0.0013)
Lasso-MRB(t=0.05)	0.869	0.859 (0.0124)	0.841 (0.0099)	0.833 (0.0089)	0.559 (0.0013)

Models are compared by their true positive rate and false positive rate (in parentheses), across 300 simulated datasets, using a significance level of α=0.01. Each column represents a scenario where a different number of null SNPs were used (e.g. 0, 45, 300, 900, or 20,000)

ISSN: 2730-6844