Regression models (using SAS 9.3) were used to investigate the main effects of childhood adversity and ADH1B-rs1229984 with each alcohol phenotype, controlling for sex, age, and former Soviet Union immigrant status, as drinking differs by these subgroups in Israel (Hasin et al., 2002; Shmulewitz et al., 2012; Spivak et al., 2007). Poisson regression models with overdispersion were used, as that distribution provided the best fit for the data (Meyers et al., 2013). Maxdrinks and AUD severity were also modeled using normal, Poisson (without overdispersion), negative binomial, and zero-inflated distributions; data best fit the overdispersed Poisson distribution best based on the likelihood and goodness-of-it indices (Akaike’s Information Criterion and the Bayesian Information Criterion). Furthermore, since the lowest value for Maxdrinks was 1, we re-ran analyses using a zero truncated model; results remained unchanged. Means were adjusted for demographic variables (sex, age, ethnicity), and mean differences indicating the difference in the mean trait value given the presence of the risk factor, and corresponding 95% confidence intervals (CIs), were computed on the additive scale using an identity link in SAS 9.3 (Spiegelman &
