Sponsored by:
The identification of causal relationships between specific genes and social, behavioral, and health outcomes is challenging due to confounding factors such as population structure and dynastic genetic effects. Sibling pairs present a useful natural experiment for the identification of causal genetic effects because, conditional on their parents' genes, a child's genes are inherited randomly via recombination. Thus, genetic differences between siblings are ignorably assigned. At present, the fixed effects model is the predominant regression specification used to estimate genetic effects using sibling pairs. Such models require four pieces of information: the genotype of both siblings and the phenotype of both siblings. We introduce a new regression specification to compare siblings and estimate direct genetic effects, which we call the phenotype differences model. Phenotype differences models require only three pieces of information: the genotype of one sibling and the phenotype of both siblings. We show that, under minor assumptions, the phenotype differences model, like the fixed effects model, provides unbiased and consistent estimates of genetic effects. After introducing the method, we apply the phenotype difference model to estimate the causal effects of ~50 polygenic scores on lifespan using asymmetrically genotyped sibling pairs in the Wisconsin Longitudinal Study.