Interaction Effects in Multiple Regression

Abstract

Multiple regression is widely used for the analysis of nonexperimental data by investigators in social work and social welfare. Most published studies test additive models in which the effects of each independent variable on the dependent variable are assumed to be constant across all levels of additional independent variables. Tests are seldom made for the presence of interacting or modifier effects. This article discusses the concept of interaction, the use of product terms to test for its presence, the problems of multicollinearity and nonlinear interaction effects and the proper use of subgroup analysis.