The present study aims at testing the effect of sample selection on the dimensionality of alcohol abuse and dependence, applying different methods. Using an empirical dataset from a general population survey, dimensionality analyses will be performed separately for a) the total sample (including the zero class) and b) a subsample of individuals with at least one endorsed criterion (excluding the zero class). It is expected that if abuse and dependence were separable constructs, both should be better identifiable when excluding the zero class. According to the literature, the effect of sample selection will be tested on the grouping of items (factor analysis) as well as the grouping of individuals (latent class analysis) and combinations of both (latent class factor analysis).
Methods
Sample