Statistical analyses were performed using standard R procedures. The significance of the replicability estimates was checked by means of a binomial test, with an expected replicability rate of 0.05 under the null hypothesis of no shared associated SNPs between Europeans and East Asians or Europeans and Africans. Similarly, the significance in the risk allele direction was checked by means of a binomial test, using a null expected ratio of 0.5. As indicated in the first section, differences in LD between replicated and non-replicated SNPs were checked by means of Mann-Whitney tests comparing the distributions of varLD scores for sliding 50-SNP windows centered on the disease-associated SNPs. The same procedure was used for the average difference in heterozygosity and distributions of OR found by “small” and “large” GWAS.
