Bayes factors for two-group comparisons in Cox regression with an application for reverse-engineering raw data from...

Area of Science:

• Biostatistics
• Medical Research Methodology
• Survival Analysis

Background:

• Cox proportional hazards regression is widely used for time-to-event data analysis in biomedical research.
• Frequentist inference is the standard approach for comparing hazard rates between experimental and control groups.
• Limitations exist in frequentist methods for interpreting evidence and resource allocation.

Purpose of the Study:

• To present a procedure for calculating Bayes factors for simple Cox models.
• To accommodate analyses using both complete datasets and summary statistics.
• To introduce the 'baymedr' R package for implementing this procedure.

Main Methods:

• Development of a procedure to compute Bayes factors for Cox proportional hazards models.
• Implementation of the procedure within the 'baymedr' R package.
• Adaptation for scenarios with full data and summary statistics.

Main Results:

• A method for computing Bayes factors in Cox models is now available.
• The 'baymedr' R package facilitates the application of this Bayesian approach.
• The procedure is applicable to both complete and summarized data.

Conclusions:

• Bayes factors offer a valuable alternative to frequentist inference in Cox model analysis.
• This Bayesian approach can address shortcomings of traditional statistical methods.
• The use of Bayes factors has the potential to optimize the use of limited research resources.