A currently common approach is to construct the polygenic score by summing the number of trait-increasing alleles across selected markers, without considering their effect sizes other than to identify the direction of association at each marker. This may be called an unweighted score, in contrast to the above approaches that estimate weights for each marker. The unweighted score may be more robust against errors in estimating the effect sizes arising from limited sample size, population heterogeneity, “winner's curse” bias, and confounding by population structure. Here a related approach is considered in which all markers are given the same absolute effect size on the standardised genotype scale. This is equivalent to the allele counting approach when all markers have the same allele frequency. When allele frequencies are heterogeneous, allele counting assumes that all markers have the same effect on the trait, whereas the present approach assumes that all markers contribute the same proportion of variance to the trait. Both models can be criticised but the present approach will allow the comparison of weighted to unweighted scores without considering the distribution of allele frequencies or their relation to the effect sizes.
