Other limitations pertain to the cross-sectional nature of this study, and sample size constraints. Given the lack of information on temporal order, we could not assess the directionality of effects. Also, despite relatively large oversamples of African Americans and Hispanics in the 2005 NAS, we did not have sufficient statistical power to examine DSM-IV alcohol dependence as an outcome in our multivariate models, nor to disaggregate analyses by gender, and by types of social consequences and dependence symptoms.
