The Impact of Unbalanced Design in Three-Level Multiple Membership Random Effects Modeling

• In Event: Methods for Complex Hierarchical Structures

Abstract

In both non-experimental and quasi-experimental designs, an equal number of observations in each group is ideal. However, unequal sample sizes are frequently encountered in practical application. This simulation study investigates the effects of unbalanced design in three-level multiple membership random effects modeling estimates. The manipulated conditions were: level-3 sample size (30, 50, and 100), level-2 sample size (10, and 20), level-1 sample size (15, and 30), multiple membership rate (10%, 20%, and 40%) and unbalanced design (balanced, moderately unbalanced, and severely unbalanced). Results showed that unbalanced design did not affect three-level multiple membership random effects modeling estimates. The relative bias of parameter estimates and root mean square errors were sensitive to the sample size, especially to the level-3 sample size.

Authors

• Hyewon Chung, Chungnam National University

• Jiseon Kim

• Hyeonjeong Jeon, National Youth Policy Institute

• Soyoung Park, Chungnam National University