Thus, the goal of the present study was to test whether distinct classes of undergraduate drinkers would emerge, based on a pattern indicative of difficulty controlling alcohol use or according to other patterns of criteria endorsement. To justify this approach, we first needed to establish evidence of measurement invariance or population heterogeneity with regard to one or more variables that may be indicative of particularly strong alcohol-related risk. We utilized a Multiple Indicator Multiple Causes or MIMIC model for this purpose. A MIMIC model with categorical variables is equivalent to an item response model where the latent factor variable is regressed on covariates to assess evidence of measurement invariance or population heterogeneity. We selected gender and age at initiation of drinking to be the covariates in this model. Gender (e.g., Harford et al., 2010) and age at initiation (e.g., Clark et al., 2005; Iacono et al., 2008) are important predictors of at-risk drinking in their own right, but are also important given their possible relationships to other predictors of at-risk drinking, such as personality traits reflecting behavioral undercontrol/disinhibition (Iacono et
