Resampling 10 000 subsets of SAGE individuals to create a sample size that was equivalent to COGA resulted in similar results. GRS showed statistically significant association with the AO-AD (P<0.05) in the reduced SAGE dataset in all 10 000 permutations. However, the reduction in sample size influenced the magnitude of P-values and only about 12% of the time, the association P-values were equal to or more significant than the original P=6.70 × 10−06 observed with the full SAGE sample.
