Bayesian Regression Analysis for Dependent Data with an Elliptical Shape.

Area of Science:

• Statistics
• Machine Learning
• Data Science

Background:

• Functional data analysis often requires models that capture complex shapes and dependencies.
• Existing methods may struggle with simultaneously modeling curved trajectories and systematic data errors.

Purpose of the Study:

• To propose a novel parametric hierarchical model for elliptical functional data.
• To accurately reconstruct curved trajectories by integrating Gaussian process priors and von Mises-Fisher distributions.
• To provide a flexible framework adaptable to higher-dimensional problems.

Main Methods:

• Development of a parametric hierarchical model incorporating Gaussian process priors for data dependencies.
• Utilizing the von Mises-Fisher distribution to model the underlying curved shape of the data.
• Implementation of Bayesian inference and Markov Chain Monte Carlo (MCMC) algorithms for parameter estimation.
• Validation using both simulated datasets and real-world examples.

Main Results:

• The proposed model successfully reconstructs curved trajectories from functional data.
• Demonstrated effectiveness in capturing systematic errors through Gaussian process priors.
• The model's ability to handle elliptical shapes using the von Mises-Fisher distribution was confirmed.
• The framework shows potential for extension to higher dimensions.

Conclusions:

• The developed model offers a robust approach for analyzing functional data with elliptical shapes and curved trajectories.
• The integration of Gaussian processes and von Mises-Fisher distributions provides a powerful tool for modeling complex data structures.
• The model's adaptability suggests broad applicability in fields requiring trajectory analysis and error modeling.