Consider a diploid individual at a particular SNP. Let the genotype at the SNP be G and the disease status be Y, where Y = 1 denotes a diseased individual and Y = 0 denotes a healthy individual. Let p = Pr(Y = 1|G). Logistic regression models are commonly used to model disease risk in GWAS [e.g. Cantor et al., 2010]. The most prominent is one in which the log-odds of disease increases (or decreases) additively by β with each copy of allele 1,
