Bayesian robust symmetric regression for medical data with heavy-tailed errors and censoring.

Area of Science:

• Biostatistics
• Medical Statistics
• Health Outcome Research

Background:

• Classical regression methods struggle with medical data due to outliers and censoring.
• Medical research frequently encounters heavy-tailed errors and incomplete observations in clinical and survival data.
• Existing methods may produce unreliable results with non-Gaussian or censored data.

Purpose of the Study:

• To develop a robust Bayesian regression model for medical data analysis.
• To address limitations of traditional methods in handling outliers and censored observations.
• To improve the reliability of statistical modeling in medical research.

Main Methods:

• Developed a Bayesian regression model incorporating symmetric error distributions (Student-t, Cauchy).
• The model explicitly handles both right and left censoring via its likelihood structure.
• Inference was conducted using Markov Chain Monte Carlo (MCMC) for uncertainty estimation.

Main Results:

• The proposed Bayesian model demonstrated superior performance over traditional methods.
• The model effectively handled noisy, censored, and non-Gaussian data in simulations and real-world applications.
• Validated through lung cancer survival analysis and hospital stay duration modeling.

Conclusions:

• The robust Bayesian symmetric regression model offers a principled framework for medical statistics.
• The approach provides improved resistance to extreme values and accounts for data censoring.
• Highlights potential for broad application in health outcome research and biostatistics.