In subsequent analyses involving NR3C1 SNPs, because genome structure differs across ethnic groups and allele frequencies vary between populations, analyses were conducted separately within EA and AA groups to avoid the potential problem of population stratification (Cardon and Palmer 2003). Frequencies of the less common alleles ranged from 5% to 44% across the 10 SNPs (see Table 2). Because individuals may carry 0, 1 or 2 copies of a given allele, two models were tested for each SNP: an additive model (SNPs were coded as 0, 1 or 2 according to the number of copies of the test allele) and a dominance model (SNPs were coded as 0 if no test allele or 1 for the presence of 1 or 2 copies of the test alleles). To test the moderation by NR3C1 variants of intervention effects, a series of 10 latent growth curve models (one for each SNP) were fitted, where the same set of covariates, intervention condition, SNP genotype, as well as their interaction term were included to predict the intercept and linear slope. A modified Bonferroni correction was
