Mobility in Three-Level Longitudinal Data: An Evaluation of the Cross-Classified Multiple Membership Latent Variable Regression Model

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

Abstract

The three-level latent variable regression (HM3-LVR) growth curve model can be used to test directional hypotheses among growth curve trajectory parameters. The cross-classified multiple-membership LVR (CCMM-LVR) model extends the HM3-LVR to handle multiple-membership data structures as is frequently found in educational datasets when there are mobile students who change schools across a study’s timeframe. In this study, we used a real data analysis and a simulation study to compare and evaluate estimation of the CCMM-LVR and HM3-LVR models. CCMM-LVR estimates were less biased (especially in scenarios with more clusters) although we found substantial bias in some parameters across conditions for both models. In the final paper, we will discuss the studies’ results as well as implications for applied researchers.

Authors

• Audrey J. Leroux, Georgia State University

• Susan Natasha Beretvas, The University of Texas - Austin