Conspicuously absent from Table III are the top-ranking TAOK3 markers from the GWAS, eight of which were genotyped but none achieving significance. The directions of genetic effect for all eight TAOK3 SNPs are the same as those obtained from the GWAS (i.e., minor alleles associated with theta ERO reductions); however, the largest effect size is 0.092 (SE 0.078) for rs7314987, approximately one-fourth of the regression coefficient estimated for the top GWAS SNP rs2784. The MAF values in the family-based sample (N = 1,095) are similar to those reported for the GWAS case-controls (N = 1,064), ranging from 0.11 to 0.13, suggesting comparable power for detecting association with the ERO phenotype. None of the markers exhibit significant population stratification or deviation from HWE. Thus, the observed discrepancy in associations across the two stages may simply represent a GWAS false positive for this particular string of markers within the TAOK3 LD block or stem from underlying differences in genetic substructure between the data sets that relate to the high-density affected family or case-control sample designs. If the smaller effect size estimate from
