We used Stata 12.0 (StataCorp 2012) to calculate descriptive statistics of the analytic sample and Mplus 7.1 for latent variable modeling. All analyses adjusted for the complex sample design of NESARC permitting the estimation of population-representative estimates with accurate standard errors. For sample descriptive statistics of sociodemographic characteristics and past-year psychiatric disorders, we calculated weighted percentages for categorical variables and means for continuous variables. We addressed missing data for the individual alcohol stigma items and for the AUD severity variable using multiple imputation with chained equations in Stata. Fewer than 6% of the analytic sample had missing data on one or more stigma items and 0.5% had missing data on AUD severity. In the imputation model we included all sociodemographic covariates. Results of models estimated with full information maximum likelihood were analogous to the results estimated from multiple imputation data.
