A Bayesian estimate of the concordance correlation coefficient with skewed data

Area of Science:

• Biostatistics
• Statistical Modeling

Background:

• The concordance correlation coefficient (CCC) is a widely used metric for assessing agreement between two measurements.
• Standard CCC estimation assumes normally distributed data, which is often violated in practice, particularly with skewed datasets.
• Existing methods for skewed data lack a comprehensive Bayesian approach and comparison.

Purpose of the Study:

• To propose and evaluate a novel Bayesian method for estimating the CCC in skewed data.
• To compare the performance of the proposed Bayesian method against existing state-of-the-art techniques.
• To highlight the advantages and flexibility of the Bayesian approach for skewed data analysis.

Main Methods:

• Development of a Bayesian framework for CCC estimation tailored for skewed data.
• Comparative analysis using simulations to assess performance under varying data conditions.
• Validation using real-life biomarker data from an electroencephalography clinical study.

Main Results:

• The proposed Bayesian method demonstrates superior performance compared to the best previously investigated method, particularly when random subject effects dominate over error variance.
• The Bayesian approach offers enhanced flexibility, accommodating missing data, confounding covariates, and replications.
• Simulation studies and real-world data analysis confirm the efficacy and advantages of the Bayesian CCC estimation.

Conclusions:

• The Bayesian method provides a robust and flexible alternative for estimating the CCC with skewed data.
• This approach enhances the reliability of agreement assessment in complex datasets, including those with missing values or covariates.
• The implementation is publicly available via the Comprehensive R Archive Network, facilitating broader adoption.