LD scores are associated with differences in allele frequencies between populations but LD score regression can still distinguish confounding from polygenicity

The LD score regression method tests whether there is an association between the LD score and allele frequency differences between cases and controls. It makes the assumption that there is no association between LD score and allele frequency differences between populations and hence that any observed association is due to a polygenic effect rather than population stratification. This assumption has not previously been tested. In comparisons between HapMap populations we observe that there is indeed an association between the LD score and allele frequency differences. However this effect is small and when we carry out simulations of large case-control samples the effect becomes negligible. We conclude that if the intercept is small then any increase in mean chi-squared does indeed reflect a polygenic effect rather than population stratification.