After selecting the appropriate LCA measurement model, we computed the measurement error associated with assigning individuals to latent classes in order to control for this error when comparing perceived alcohol stigma across the latent classes. In the statistical literature, this approach is referred to as 3-step analysis (Vermunt 2010). We used a Wald test to compare adjusted mean perceived alcohol stigma scores across the latent classes. Means were adjusted by regressing perceived alcohol stigma on sociodemographic characteristics and AUD severity. We computed Cohen’s d (Cohen 1988) as an effect size measure for differences in adjusted means. A Cohen’s d of 0.2, 0.5, and 0.80 indicate small, medium, and large differences, respectively.
