To incorporate the impact of demographic factors, the outcome variable used in the twin models is the residual of PYDD after partialing the effects of demographic covariates based on the baseline multiple regression model for each sex (see Fig. 1). Parallel to the individual-level analyses, model building proceeded by adding the main effects of drinking onset and SLE, followed by their interaction term. As in multiple regression, the resulting regression coefficients (in Table 4) refer to the change in PYDD (in standard deviation units) for each standard deviation change in the covariate. We evaluated nested models by comparing the difference in variance of PYDD attributable to genetic and environmental sources in the baseline (i.e., demographics-only) model upon addition of each covariate. For instance, a reduction in genetic variance of PYDD after adding the SLE term would suggest that the influence of SLE on PYDD is mediated by genetic effects. We report results for the 3 types of SLE, analyzed separately in men and women.
