The UK Biobank is among the 11 cohorts included in the Polygenic Index Repository. Because of its large sample size (see Table 2), the UK Biobank contributes substantially to the available sample for the GWAS for many phenotypes. We therefore did not want to exclude the entire UK Biobank from the GWASs used to create the PGIs. Instead, we split the UK Biobank sample into three equal-sized partitions. We ran three 1/3-sample GWASs for each phenotype. To create the PGI for each partition, we included results from the other two partitions in the meta-analysis. Consequently, researchers can conduct analyses of a PGI in any one of the partitions and obtain unbiased results. However, we caution researchers against conducting analyses in two or three of the partitions and meta-analyzing across partitions; because the other partitions are used to create the PGI, the results obtained across different partitions (although individually unbiased) will be correlated. Meta-analysis standard errors will therefore be anticonservative, and this bias can be substantial (see Methods). Therefore, to maximize the usefulness of our PGIs for research involving related individuals
