Graphical models for mean and covariance of multivariate longitudinal data

Area of Science:

• Statistics
• Biostatistics
• Data Science

Background:

• Analyzing multivariate longitudinal data is crucial for understanding correlated outcomes.
• Complex covariance structures pose a significant challenge in such analyses.
• Visualizing dependence patterns among longitudinal outcomes is often difficult.

Purpose of the Study:

• To demonstrate the utility of graphical techniques (profile plots, multivariate regressograms) in mean-covariance modeling.
• To introduce the MLGM R package for visualizing and modeling multivariate longitudinal data.
• To address the lack of readily available data-driven tools for dependence pattern visualization.

Main Methods:

• Joint mean-covariance modeling of multivariate longitudinal data.
• Application of graphical techniques: profile plots and multivariate regressograms.
• Development and utilization of the MLGM R package for analysis.

Main Results:

• Graphical methods effectively aid in developing mean and covariance models.
• The MLGM package facilitates visualization and modeling of complex data patterns.
• The proposed approach shows good performance in real-world data and simulations.

Conclusions:

• Graphical techniques are valuable for understanding multivariate longitudinal data.
• The MLGM R package provides a practical tool for researchers.
• The developed methods offer a robust approach for analyzing complex longitudinal dependencies.