Fig. 5.

Multilevel parametrization of the ACE model with Region as a third level, which loads on the region variable REG F on the family level, on which the observed variable Y is regressed with its coefficient estimating the effect of region. Sex and age are covariates. We compared the model where we included PC1 as a covariate with the model where we did not include PC1 as a covariate (we tested PC1, PC2, and PC3, but depict only PC1 to avoid clutter, and because our final model included only PC1). Note that the dummy REG F latent variable serves as a placeholder to stress the nesting of families in regions, but technically, it is not needed