The Impact of Ignoring Individual Mobility Across Clusters in Multilevel Discrete-Time Survival Analysis

• In Event: Methods for Complex Hierarchical Structures

Abstract

Little work has explored the use of a multilevel discrete-time survival analysis in the presence of data that contains individual mobility across clusters. This simulation study examines bias in parameter estimates when mobility is incorrectly modeled in a conventional multilevel discrete-time survival model as compared to modeling mobility using both a multiple membership discrete-time survival model and a cross-classified discrete-time survival model. This study manipulated the overall mobility rate, the cluster size, and the within-cluster sample size. The findings suggest that while there is substantial relative parameter bias across all models, the multiple membership discrete-time survival model performs best in the recovery of model parameters, specifically in regard to the cluster-level slope and estimates of the variance component.

Authors

• Christopher Cappelli, Georgia Institute of Technology

• Audrey J. Leroux, Georgia State University

• Katherine E. Masyn, Georgia State University