Latent variable models of alcohol-related constructs

This study examines the improvement in validity coefficients and structural relationships derived from formally modeling latent variables of adverse alcohol-related constructs. Results of a confirmatory factor analysis of responses from 8, 755 participants in the 1988 National Household Survey of Drug Abuse support a three-factor model of alcohol abuse, dependence, and adverse consequences. A comparison of construct inter-correlations shows latent variables to be superior to measurement approaches using unweighted sums, quantity × frequency calculation, and single-item measurement