Extending multivariate- t linear mixed models for multiple longitudinal data with censored responses and heavy tails

Area of Science:

• Biostatistics
• Statistical Modeling
• Longitudinal Data Analysis

Background:

• Longitudinal data analysis presents challenges including multiple response series, irregular sampling times, and data censoring.
• Outliers and heavy-tailed noise can further complicate the accurate modeling of such complex datasets.

Purpose of the Study:

• To develop a statistical model capable of simultaneously addressing multiple features inherent in complex longitudinal data.
• To introduce the multivariate-t linear mixed model with censored responses (MtLMMC) for robust data analysis.

Main Methods:

• Formulation of the multivariate-t linear mixed model with censored responses (MtLMMC).
• Development of an efficient Expectation Conditional Maximization (ECM) algorithm for maximum likelihood estimation.
• Utilizing truncated multivariate-t distributions and auxiliary permutation matrices for computational efficiency.

Main Results:

• The proposed MtLMMC effectively models longitudinal data with censored responses, outliers, and heavy-tailed noise.
• The ECME algorithm provides efficient parameter estimation for the developed model.
• Demonstrated applicability through simulation studies and a real-world HIV/AIDS dataset analysis.

Conclusions:

• The MtLMMC provides a powerful and flexible framework for analyzing complex longitudinal data with inherent challenges.
• The developed methodology offers improved accuracy and efficiency in statistical modeling for such data types.