After establishing the best-fitting CFA model, we used SEM to examine the construct validity of the latent stigma factors, by testing associations between PAS and social support and social network involvement. First we specified unadjusted and adjusted main-effects models to examine these relationships among lifetime drinkers and separately by alcohol-affected status. Next, we ran models among lifetime drinkers while including an interaction between latent PAS and alcohol-affected status, using the Mplus latent moderated structural equations approach with the maximum likelihood estimator. Lifetime abstainers were excluded from these analyses, as described later, due to the poorer model fit of the PDD in that subgroup.
