The Impact of Ignoring a Higher Level of Nesting in Cluster-Randomized Experiments: A Simulation Study

• In Event: Power and Design, Model Misspecification, and Multilevel Modeling

Abstract

Hierarchical linear modeling (HLM) is frequently employed to analyze multilevel data from cluster-randomized trials (e.g., Russell, O'Dwyer, and Miranda, 2009). However, empirical research using HLM in this context frequently ignores levels of nesting above the level at which randomization occurs; this often-ignored higher level may be statistically or conceptually important to the analysis. The present study will investigate the impact of ignoring a higher level of nesting in the context of cluster-randomized trials. Four-level longitudinal data will be simulated and analyzed under both the four-level model and the three-level model which ignores the highest level of nesting structure. Implications of the findings will be discussed.

Authors

• Ryan Glaman, University of North Texas

• Qi Chen, University of North Texas