The Effect of Correlated Clusters on Parameter Estimates in Multiple Membership Models

• In Event: 79.046 - Studies Addressing Issues Related to the Modeling of Multiple Membership Data

Abstract

Multiple membership models provide researchers interested in examining mobile populations tools to estimate parameters across hierarchical structures with minimal bias. These models have received quite a bit of attention in the literature in recent years, especially as innovations lead to increased computation power and improved estimation methods. Given that the development of these models is relatively new, certain model assumptions have yet to be assessed with rigorous simulation studies in the literature. The present study investigates the impact of violations of non-independence at the highest level of clustering on parameter estimates and aims to use these results to develop guidelines for modeling mobility when explicitly including higher-level clustering units is infeasible or undesirable.

Authors

• Yi Feng, University of Maryland - College Park

• Tessa Johnson, University of Maryland - College Park

• Laura M. Stapleton, University of Maryland

• Yating Zheng, University of Maryland - College Park