glmmPen: High Dimensional Penalized Generalized Linear Mixed Models

Area of Science:

• Statistics
• Computational Biology
• Biostatistics

Background:

• Generalized linear mixed models (GLMMs) are essential for analyzing correlated, non-Gaussian data.
• Accurate selection of fixed and random effects is crucial to prevent bias in GLMMs.
• Previous methods for joint effect selection were limited to lower-dimensional problems.

Purpose of the Study:

• Introduce the R package glmmPen for high-dimensional GLMMs.
• Develop a penalized modeling framework for joint fixed and random effects selection.
• Provide an efficient computational algorithm for parameter estimation.

Main Methods:

• Utilize a penalized generalized linear mixed model framework.
• Employ a Monte Carlo expectation conditional minimization (MCECM) algorithm for parameter estimation.
• Leverage Stan and RcppArmadillo for computational efficiency in the glmmPen package.

Main Results:

• The glmmPen package facilitates joint selection of fixed and random effects in high-dimensional GLMMs.
• The MCECM algorithm provides efficient parameter estimation.
• Simulations demonstrate good performance in selecting both fixed and random effects.

Conclusions:

• glmmPen offers a novel solution for high-dimensional GLMMs, addressing limitations in effect selection.
• The package supports Binomial, Gaussian, and Poisson families with various penalty functions.
• This method improves the accuracy and reliability of GLMM analysis in complex research areas.