Assessing the Significance of Random Slope Effects in Multilevel Models

• In Event: Advances in Multilevel Modeling

Abstract

The multilevel model is commonly used in educational research, and a random slope effect (where the slope between a predictor and outcome is allowed to vary randomly by group membership) can be an important effect to assess within these models. For example, the random slope effect is commonly used to examine individual growth trajectories in longitudinal models and for examining treatment-by-site variance in multisite randomized trials. However, it is not clear how to assess the significance of these effects. The present study considers four methods for doing so: two likelihood ratio tests, and two information criteria approaches. The results indicate that all four methods show desirable properties. Implications and suggestions for applied researchers are provided.

Author

• Julie A. Lorah, Clemson University
  Presenting Author