An Efficient Approach to Screening Epigenome-Wide Data

Table 2

Simulation results for selecting important variables among 2,000 candidates with surrogate variables excluded.

Statistics	Bon	FDR	TT	Bon	FDR	TT
Random error normally distributed
# incorrect	10	51	9	99	114	90
Sensitivity	0.011	0.054	0.100	0.014	0.167	0.102
Specificity	1	0.979	1	1	0.984	1
# incorrect	197	182	180	394	310	359
Sensitivity	0.015	0.232	0.102	0.015	0.288	0.104
Specificity	1	0.984	1	1	0.984	1
Random error distributed (df = 1)
# incorrect	10	38	9	98	103	91
Sensitivity	0.018	0.053	0.095	0.015	0.158	0.096
Specificity	1	0.986	1	1	0.99	1
# incorrect	197	176	181	394	314	363
Sensitivity	0.014	0.217	0.094	0.014	0.273	0.095
Specificity	1	0.989	1	1	0.986	1

Bon: Bonferroni, FDR: false discovery rate, and TT: training and testing.