Test of a Drug Use Causal Model Using Asymptotically Distribution Free Methods

Abstract

Previous statistical comparisons of two models for adolescent drug abuse are reexamined using new statistical estimation methods in causal modeling not requiring assumptions about normally distributed variables. Browne's asymptotically distribution free method shows that the models fit even better than assumed in the initial work, and that comparable inferences would be drawn about the significance of individual parameters and model comparisons when the assumption that the data are distributed in a multivariate normal manner is relaxed.