Testing and Interpreting Interaction Effects in Multilevel Models

• In Event: Longitudinal Data, Model Evaluation, and Multilevel Models

Abstract

Our purpose is to demonstrate the analysis and interpretation of cross-level interactions in multilevel models and describe results of a study of mathematics achievement growth for students with specific learning disabilities (SLDs) and general education (GE) students. A second purpose was to test whether student socio-demographic characteristics interact with the growth parameters or group differences in the longitudinal model. Many multilevel growth studies do not fully test cross-level interactions nor include within-level interaction terms even when research hypotheses concern moderating effects. Using two statewide samples of approximately 80,000 students over Grades 3-7, we describe and demonstrate methods for exploring and testing multilevel interaction effects. We discuss the importance of explicitly testing interactions as well as implications for future research.

Authors

• Joseph J. Stevens, University of Oregon

• Ann C Schulte, Arizona State University - Tempe