In GS:SFHS, the associations between PGS and MDD, rMDD, fMDD, and mMDD were tested using a mixed linear model, covarying for the first 20 PCs to account for population stratification. Prior to this analysis, the requisite number of PCs was established using a stepwise linear regression approach, adding one PC at a time, and using a likelihood-ratio test (LRT), the output of which was assessed against a mixed 0.5(χ2) + 0.5(0) distribution56. An additive genetic component was fitted as a random effect to account for the increased relatedness within GS:SFHS. To ensure that common environment was adequately modeled, models incorporating shared parent-offspring, sibling, and spousal environmental components as additional random effects were tested using a stepwise LRT approach, however no environmental component improved model fit. Further details of mixed linear model selection are provided in the Supplementary Methods. F-statistics, degrees of freedom, effect sizes, Z-scores and P-values were derived using the Wald Conditional F-test57, in ASReml-R58.
