Exploratory factor analysis (EFA; Fabrigar et al., 1999) was used to determine the number of factors required to explain the covariance across the 2 alcohol abuse (i.e. excluding hazardous use and legal problems) and 7 dependence criteria for all participants and in groups defined by DUI status and gender (men and women with DUIs; men and women with no DUIs) before proceeding with an item response theory (IRT; Muthén and Lehman, 1985) analysis. The EFA results supported the unidimensionality of the criteria and a confirmatory one factor model (using confirmatory factor analysis or CFA) was used for all subsequent models. Parameters from CFA can be converted into a 2 parameter (2P) logistic item response model (Muthén and Lehman, 1985) characterized by discrimination and difficulty. Discrimination represents the ability of a criterion to distinguish individuals at high risk for AUDs from those at relatively lower risk. Difficulty reflects the location along the liability continuum where the criterion has a .50 probability of endorsement. It thus represents the threshold on the alcohol severity continuum where the criterion becomes more likely to be endorsed than not.
