Variable selection for joint models of multivariate skew-normal longitudinal and survival data.

Area of Science:

• Biostatistics
• Clinical Trials
• Longitudinal Data Analysis
• Survival Analysis

Background:

• Existing joint models for longitudinal and survival data often assume normality, limiting their applicability.
• There is a need for methods that can handle non-normality in longitudinal outcomes while performing variable selection.
• Simultaneous parameter estimation and variable selection are crucial for robustly modeling complex biomedical data.

Purpose of the Study:

• To develop a novel joint modeling framework for multivariate skew-normal longitudinal and survival data.
• To incorporate simultaneous parameter estimation and variable selection within this framework.
• To identify significant covariates and trajectory functions, and detect deviations from normality in longitudinal data.

Main Methods:

• Utilized penalized splines for estimating the log baseline hazard function.
• Employed the rectangle integral method to approximate the conditional survival function.
• Developed a Monte Carlo expectation-maximization algorithm for parameter estimation and a one-step sparse estimation procedure for variable selection.

Main Results:

• The proposed method effectively performs simultaneous parameter estimation and variable selection in joint models.
• It successfully identifies significant covariates and trajectory functions, enhancing model interpretability.
• The methodology can detect departures from normality in longitudinal data, offering greater flexibility.

Conclusions:

• The novel joint modeling approach provides a powerful tool for analyzing complex longitudinal and survival data with non-normality.
• The integrated variable selection enhances model parsimony and identifies key predictive factors.
• The method is validated through simulation studies and a real-world clinical trial example.