A0328
Title: Innovative regularization with the EM algorithm in linear mixed models for variable selection
Authors: Daniela Oliveira - Federal University of Sao Joao Del Rei (Brazil) [presenting]
Fernanda Schumacher - The Ohio State University (United States)
Victor Hugo Lachos Davila - University of Connecticut (United States)
Abstract: The EM algorithm has seen little application in high-dimensional regularization for linear mixed-effects models. An innovative approach combines the EM algorithm with the efficient cv.glmnet package for tuning the lambda parameter and the glmnet package for Lasso variable selection in such models. The method is thoroughly evaluated through simulation studies and a riboflavin application, comparing its performance to existing R packages: glmmLasso and splmm. Results show that the proposed approach is robust and effective, even when the number of predictors exceeds the number of observations or when there are missing values in the response variable.
