Comparing the Prediction Performance of MERF and glmmLasso for Longitudinal Data

• In Event: Fresh Perspectives on Estimating Complex Multilevel Models: Counts, Mixtures, Interactions, and Machine Learning

Abstract

Mixed-effects random forest (MERF) and generalized linear mixed models with Lasso regularization (glmmLasso) have been introduced to address the complexity of longitudinal data structures. This simulation study aims to compare the prediction performance of MERF and glmmLasso under various conditions, including sample size (500, 1,000, 2,000, and 4,000), the number of measurement occasions (3, 4, and 6), and the number of predictors (25, 50, and 100). In general, glmmLasso outperforms MERF in terms of prediction performance. As the number of predictors increases, the root mean square error (RMSE) decreases for glmmLasso and increases for MERF. For both glmmLasso and MERF, the RMSE increases as the number of measurement occasions increases, and the RMSE decreases as the sample size decreases.

Authors

• Hyewon Chung, Chungnam National University

• Jiwon Lee, Chungnam National University

• Eunah Jang, Chungnam National University