Models were tested individually in each cohort and then combined using a random-effects inverse variance-weighted meta-analysis of the PGS effect estimates from each model, using the metafor package. Results from FTC were not meta-analyzed due to the difference in age and available measures, but were compared qualitatively with the primary LGC meta-analysis results. In FTC, only IS models were selected as many participants only had data for 3 or fewer measurement occasions and the quadratic effect estimates were unreliable. We applied a Bonferroni correction for 56 PGS-outcome pairs, resulting in an adjusted alpha=.05/56=.0009.
