Specifically, we combined GWAS results weights from Karlson Linnér et al. (in press) with results from a 10-fold GWAS method of a latent externalizing factor constructed by a phenotypic externalizing factor that corresponded with the seven indicator phenotypes in the discovery multivariate analyses (Karlsson Linnér et al. in press) in the COGA sample of African ancestry individuals (including all COGA African ancestry individuals, not limiting to the subset of African adolescents in the current study; n = 3379). Thus, for our African ancestry sample, we calculated two versions of the externalizing polygenic scores. The externalizing polygenic score that accounted for the largest amount of variance in adolescent externalizing behavior scores in the African ancestry sample was advanced to subsequent analyses of that sample, per our pre-registered analytic plan. We recognize the limitations of this approach and understand that it is less than ideal; however, we believe it is an ethical imperative to include African Americans in genetic studies like this one.
