![IconGems-Print](images/IconGems-Print.png)
519 – Sparse Statistical Learning
Variable Selection for a Mixture of Linear Mixed Effects Models
Yian Zhang
Lei Yang
Yongzhao Shao
New York University School of Medicine
Linear mixed effects models have been widely used in medical studies and other applications to deal with repeated-measure or clustered data. In reality, these data can be a mixture of different groups of subjects. Then, it is preferred to fit the group-specific model if the group indicator is available so that existing variable selection methods for linear mixed effects models can be used. In this study, we consider a situation that the data involve two groups of subjects but the grouping information is only partially known. We construct a finite mixture of linear mixed effects models via joint modeling of repeated measures using linear mixed effects models and missing group status using a logistic regression. We propose a penalized likelihood method with SCAD penalty function using EM algorithm for variable selection in this mixture model. BIC is applied to determine the tuning parameters and covariates. Simulation studies are used to demonstrate the performance of the proposed method.