Bayesian semiparametric nonlinear mixed-effects joint models for data with skewness, missing responses, and...

Area of Science:

• Biostatistics
• Longitudinal Data Analysis
• Statistical Modeling

Background:

• Semiparametric nonlinear mixed-effects (SNLME) models commonly assume normal error distributions.
• Normality assumptions can obscure subject-specific variations and complicate analyses with skewness, missing data, or measurement error.
• Existing methods often address these issues individually, not simultaneously.

Purpose of the Study:

• To develop a flexible Bayesian SNLME model capable of simultaneously addressing skewness, missing responses, and covariate measurement error.
• To provide a robust framework for analyzing complex longitudinal data with these challenges.
• To illustrate the model's application on a real-world AIDS dataset.

Main Methods:

• Jointly modeling response and covariate processes within a Bayesian SNLME framework.
• Employing flexible distributional specifications to accommodate non-normal error structures.
• Utilizing a Bayesian approach to handle complex dependencies and uncertainties.

Main Results:

• The proposed Bayesian SNLME model effectively accommodates skewness, non-ignorable missingness, and covariate measurement error.
• The model provides a unified approach to handle multiple data complexities in longitudinal studies.
• Demonstrated utility through application to an AIDS data set, comparing various model specifications.

Conclusions:

• The developed Bayesian SNLME model offers a powerful tool for analyzing longitudinal data with simultaneous skewness, missingness, and measurement error.
• This approach enhances the accuracy and interpretability of statistical inferences in complex datasets.
• The methodology is broadly applicable to various fields dealing with similar data challenges.