Detecting and explicating interactions in categorical data.

Detecting and explicating interactions in categorical data analyses using cross tabulation and the [chi] 2 statistic can provide salient tests of hypotheses concerning the relationship between two variables measured at the nominal or ordinal levels. For example, researchers usually employ categorical analysis when they are interested in whether members of one group (e.g., males vs. females) differ in the proportion falling into two or more levels of a dependent variable (e.g., in favor of or opposed to sex education in public schools). In this case, the data can be expressed as a two-way table and hypotheses tested with the [chi] 2 statistic. Interpretation of this simplest of two-way tables is straightforward. However, research questions are often more complex than this simple example both in the number of predictor variables and the number of levels of each variable. Researchers typically include other predictor variables (e.g., race, academic status, marital status) to gain a better understanding of more complex relationships among predictors and outcomes. In addition, researchers often employ measures that have more than two levels (e.g., income, race, treatment type, academic status), and they often choose to combine levels in one or more variables to simplify the analyses, meet assumptions, or clarify the results.