To account for familial confounding, we used GCTA on four samples. (1) The best way to estimate the random effect of SNPs (the aggregate effect of common SNPs on the phenotype) is with a large sample of unrelated individuals (Yang et al., 2010). The largest such sample in the present study consists of all genetically unrelated parents (n = 3542), under a simplifying assumption of no assortative mating. To determine genetic relatedness in this parent sample we excluded one individual of every pair of individuals who had a genetic relatedness of ≥ .025 as calculated by the GRM produced by GCTA on the full sample. (2) To help inform the biometric twin heritability estimates, we also estimated the random effect of SNPs on an unrelated sample of the youth offspring (n = 1784), including one member from each twin and sibling pair, as well as all adopted youth. Ideally, this would provide an estimate of the aggregate effect of common SNPs in the offspring youth sample and would be comparable to that produced by the unrelated parent sample. (3) We
