Set p(j) = μ(j)/2, an estimate of the underlying allele frequency (autosomal data). Then each entry in the resulting matrix is Equation 3 is a normalization step, which is motivated by the fact that the frequency change of a SNP due to genetic drift occurs at a rate proportional to per generation. It also normalizes (at least if the data is in Hardy-Weinberg equilibrium) each data column to have the same variance. We note that Nicholson et al. use the same normalization, and motivate it similarly [18].
