Efficient Particle Smoothing for Bayesian Inference in Dynamic Survival Models

TL;DR

This paper introduces an efficient particle smoothing algorithm for Bayesian inference in dynamic survival models, significantly improving sampling efficiency and scalability over traditional MCMC methods.

Contribution

It develops a novel particle smoothing approach tailored for piecewise exponential hazard models, enhancing computational efficiency and scalability in survival analysis.

Findings

Over two orders of magnitude increase in effective sample size compared to MCMC

Algorithm scales well with high-dimensional and large datasets

Demonstrated effectiveness on both simulated and real data

Abstract

This article proposes an efficient Bayesian inference for piecewise exponential hazard (PEH) models, which allow the effect of a covariate on the survival time to vary over time. The proposed inference methodology is based on a particle smoothing (PS) algorithm that depends on three particle filters. Efficient proposal (importance) distributions for the particle filters tailored to the nature of survival data and PEH models are developed using the Laplace approximation of the posterior distribution and linear Bayes theory. The algorithm is applied to both simulated and real data, and the results show that it generates an effective sample size that is more than two orders of magnitude larger than a state-of-the-art MCMC sampler for the same computing time, and scales well in high-dimensional and relatively large data.