Based on recommendations by Dudbridge,29 that replication of polygenic scores is optimized when the size of the discovery and test samples is approximately equal; we performed 10 000 iterations in which we randomly resampled 1788 individuals from the pool of 2336 subjects in SAGE. Cox proportional hazard models were fit to each randomly drawn sample to examine whether the magnitude of the association was modified by selection of a comparatively sized test sample. Similar analyses were not conducted in OZ-ALC as random selection of subsets of individuals nested in pedigrees would not be representative of the sampling design nor would selection of subsets of whole pedigrees allow for adequate numbers of individuals with AO-AD.
