ABSTRACT
This chapter introduces some statistical approaches to testing for interaction between two predictor variables. When the predictor variables are all categorical, interaction is usually tested with factorial analysis of variance. The researcher may be interested in whether the quantitative variable is related to the outcome variable differently in the two groups or whether the average difference between the groups on the outcome variable depends on the values of the quantitative variable. Consider a simple hypothetical study similar conceptually to Monahan and Lannuitti. The data from this hypothetical study will be used throughout this section, and the details about the data file used to generate the analyses here can be found in Appendix E9 on the CD. A variable is mean centered by subtracting the mean of the variable from each case. One argument advanced for mean centering in moderated multiple regression is that it reduces multicollinearity between the product and the constituent terms of the interaction.
