Using Association Models in Research on Drug Use: Some Examples

Abstract

This article shows how association models can be used to guide and simplify analysis of cross-classifications involving ordinal variables. To illustrate, we analyze three crosstables of high school seniors pertaining to the association between certain characteristics (e.g., attitude toward school) and drug use. These models can be used to answer several questions that arise naturally in analyses of such data: 1) Is there association between the variables? 2) If there is association, can it be summarized in terms of a single index, or must a set of indexes be used? 3) How can association be broken up into component sources? 4) Can the ordinal variables be scaled in a meaningful fashion using just the information available in the cross-classification? 5) How can association between two variables be studied in an across-group (e.g., across-time) context?