be partially due to uncorrected population stratification. The dependence of correlation estimates on discovery GWAS is further illustrated in 4c, in which the point estimate for correlation between height and East Asian GWAS based polygenic scores for height is negative (r = −.11, p = .643). Power in discovery GWAS is also relevant, and greater confidence should be assigned to the results in 4a and 4b because both European ancestry discovery GWAS were adequately powered to detect hundreds of height loci, whereas the East Asian height GWAS was only adequately powered to detect 17 loci. Finally, methodological choices in polygenic score construction (see Fig. 3) must also be considered. The shape, dispersion, and even the ordering of distributions of polygenic scores for different 1000Genomes populations depends on polygenic score construction parameters, and would also necessarily result in different correlations with population phenotypes, and this is one additional reason why differences in polygenic scores among populations cannot be naively interpreted.
