Sibling Models Can Test Causal Claims without Experiments: Applications for Psychology

Randomized experiments are the "gold standard" for inferring causation and are designed to collect covariate information. Yet, many questions cannot be answered with experiments practically or ethically. Often, potential confounds are controlled statistically as covariates in quasi-experimental designs. The typical use of covariates does not control for many systematically confounded gene-and-environmental effects. Poverty, health, and individual differences all covary with gene-and-environmental effects, so much so that using covariates can create bias. We advocate for using genetically informed designs, which strengthen causal inference by controlling for major genetic and environmental confounds even in the absence of random assignment. We adapted the reciprocal standard dyad model into an analytic method to facilitate sibling comparisons. Differences between kin pairs explicitly distinguish within-family variance from between-family and control for all background variance linked to gene-and-environmental differences. We present four vignettes with individual differences and health outcomes. Although all four illustrations found significant associations when using covariate-based approaches, results diverged after addressing familial confounding with the discordant-kinship model. This divergence highlights the importance of considering familial influences in psychological research, demonstrating the versatility and efficacy of our method in different contexts.