Comparing Two Multiple Membership Weights for Cross-Classified Multiple Membership Multilevel Modeling

• In Event: Advanced Techniques in Multilevel Modeling and Treatment Effect Analysis

Abstract

This simulation study was designed to compare two different weight assignment techniques for multiple membership data in cross-classified multiple membership multilevel modeling estimates. Two multiple membership weights were compared: equal and unequal weights under manipulated conditions. The intra-unit class correlation (0.1, 0.2, and 0.3), number of groups (20, 30, 50 and 100), average group size (10, 20 and 40), cross-classification rate (20% and 40%), and multiple membership rate 10% and 20%) were varied. In general, no substantial differences were observed in the relative biases of the parameter estimates, the credible interval rates, the root mean square errors or the degrees of model fit across all the conditions manipulated in the current study.

Author

• Hyewon Chung, Chungnam National University