Table 1.

Power Analysis.

	Power (%)
Noise	Model 1	Model 2	Model 3
None	100	100	99
GE	100	100	96
MS	100	100	97
PC	89	100	49
GH	100	100	82
GE+GH+PC+MS	51	71	34

Power of HFCC to detect a two-locus interaction at an alpha level of 0.01 under 3 genetic epistasis models and different sources of noise: no noise (None), genotyping error (GE), missing data (MS), phenocopy (PC), genetic heterogeneity (GH), and all sources of noise simultenously (GE+GH+PC+MS). For the datasets with genetic heterogeneity, we reported the power to detect either of the two simulated two-locus effects. Model 1 is a heterozygote-homozygote risk (i.e., risk genotypes ABCC, ABDD, AACD and BBCD); Model 2 represents a "2 and only 2"-allele risk (i.e., A and C) from any of two loci (i.e., risk genotypes AADD, ABCD, and BBCC); Model 3 represents a protective double-heterozygote (i.e., protective genotype ABCD).