Testing Interaction Effects in Structural Equation Models
American Educational Research Association Annual Meeting,
Research questions that test the interaction between either categorical variables or multiplicative expressions are commonly found in multiple regression and analysis of variance applications. In contract, research questions posing an interaction between observed variables in a path analytic model are not commonly found in the literature. This is due in part to some of the issues and problems that arise in relational path models. Path models with latent variables also pose problems for the researcher that make the inclusion of interaction tests difficult. This paper further explores the categorical, group, or multi-sample approach and the continuous or multiplicative approach to testing interaction effects in structural equation models that use latent variables. Some software programs have provided a way to test simple interaction effects among latent variables. Examples are included. The developers of structural equation modeling software programs are encouraged to develop procedures for testing interaction effects. Until then, the testing of interaction effects in structural equation models will not be easy to perform. Of the two approaches discussed, multi-sample and multiplicative, the multi-sample approach is generally an easier technique to use for testing the equality of coefficients. An appendix contains a multi-sample interaction example and a LISREL 8 program example. An additional attachment contains other software examples and the outputs of some programs. One table and one figure illustrate the discussion. (Contains 29 references.) (Author/SLD)
Schumacker, R.E. & Rigdon, E. (1995). Testing Interaction Effects in Structural Equation Models. Presented at American Educational Research Association Annual Meeting 1995.