Predictive Assessment and Comparison of Bayesian Survival Models for Cancer Recurrence

TL;DR

This paper introduces new methods for assessing and comparing Bayesian survival models in cancer recurrence studies, addressing challenges posed by complex data features like censored times and time-dependent effects.

Contribution

It provides targeted recommendations for predictive assessment and comparison of Bayesian survival models, along with open source code for practical application.

Findings

New assessment methods for complex survival data

Guidelines applicable to various models and scenarios

Open source tools for implementation

Abstract

Complex data features, such as unmodelled censored event times and variables with time-dependent effects, are common in cancer recurrence studies and pose challenges for Bayesian survival modelling. Current methodologies for predictive model checking and comparison often fail to adequately address these features. This paper bridges that gap by introducing new, targeted recommendations for predictive assessment and comparison of Bayesian survival models. Our recommendations cover a variety of different scenarios and models. Accompanying code together with our implementations to open source software help in replicating the results and applying our recommendations in practice.