Bayesian Variable Selection and Estimation in Semiparametric Simplex Mixed-Effects Models with Longitudinal...

Area of Science:

• Statistics
• Biostatistics
• Longitudinal Data Analysis

Background:

• Standard mixed-effects models assume normal distribution for random effects, which is often violated in skewed or multimodal longitudinal data.
• Violated normality assumptions can lead to inaccurate parameter estimates and unreliable covariate selection in longitudinal studies.

Purpose of the Study:

• To develop a robust statistical framework for semiparametric simplex mixed-effects models that accommodates non-normal random effects.
• To enhance the estimation of model parameters and the selection of significant covariates in longitudinal data analysis.

Main Methods:

• Adoption of the centered Dirichlet process mixture model (CDPMM) to flexibly model non-normal random effects.
• Extension of the Bayesian Lasso (BLasso) method combined with block Gibbs sampler and Metropolis-Hastings algorithm for simultaneous parameter estimation and covariate selection.

Main Results:

• The proposed CDPMM-based semiparametric simplex mixed-effects model effectively handles skewed and multimodal longitudinal data.
• The extended BLasso approach demonstrated accurate estimation of parameters and successful identification of important covariates.

Conclusions:

• The developed methodology provides a powerful tool for analyzing complex longitudinal data where normality assumptions are not met.
• This approach offers improved statistical inference for semiparametric simplex mixed-effects models, applicable in various scientific fields.