We used an iterative pruning approach to identify independent IMPACT annotations. For each trait, we ranked all 707 IMPACT annotations by their τ* significance values. Then, we selected the lead annotation, removed all annotations correlated with Pearson r > 0.5 and selected the next lead annotation, and so on. For each trait, we regressed the EUR τ* on the EAS τ* using Deming regression, to account for standard errors, with the R function deming from the package deming. We tested the null hypothesis that the slope is equal to 1.