Fig. 1: A simplified path diagram for the structural equation model.

Note: Represents one example family with each parent having one offspring; the full model includes up to two offspring per parent. The inclusion of both offspring of half-sibling and full-sibling parents allowed us to decompose the variance of genetic (A) and environmental (E) factors. Contribution of shared environment (C) are also estimated, but left out of the figure. Y₁₁ and Y₂₁ represent two siblings in the parent generation. Y₁₂ and Y₂₂ represent two cousins in the offspring generation; A₁₁ and A₂₁ represent the parental additive genetic sources of variance, and g represents the genetic similarity between the two (i.e., 0.50 for full-siblings and 0.25 for maternal half-siblings); E₁₁ and E₂₁ represent the unique environmental contribution to the variance in the parental phenotype. A₁₂ and A₂₂ represent the offspring additive genetic sources of variance, and 0.25g represents the genetic similarity between the two; E₁₂ and E₂₂ represent the unique environmental contribution to the variance in offspring phenotype; r_g is the genetic correlation between the parental and offspring phenotype, thus 0.50r_g is the correlation between the parental and offspring phenotype due to shared genetics; similarly, 0.50gr_g is the correlation between uncle/aunt and niece/nephew due to shared genetics. Parental and offspring may have different proportion of variance explained by A and E, as seen by having different path coefficients (e.g., a_p and a_o). Finally, the direct, phenotypic intergenerational association is modeled by β, where the variance in parental phenotype, regardless of source, may directly influence the variance in offspring phenotype. A description of the model can be found in Supplementary Appendix 1 and in Kuja-Halkola et al.³².