Non-centering for discrete-valued state transition models: an application to ESBL-producing E. coli transmission in Malawi

2025/04/16 by Neill, James, Lester, Rebecca, Bakali, Winnie +4
#FOS: Computer and information sciences #Methodology (stat.ME)

paper · doi:10.48550/arxiv.2504.11836

Abstract

Infectious disease transmission is often modelled by discrete-valued stochastic state-transition processes. Due to a lack of complete data, Bayesian inference for these models often relies on data-augmentation techniques. These techniques are often inefficient or time consuming to implement. We introduce a novel data-augmentation Markov chain Monte Carlo method for discrete-time individual-based epidemic models, which we call the Rippler algorithm. This method uses the transmission model in the proposal step of the Metropolis-Hastings algorithm, rather than in the accept-reject step. We test the Rippler algorithm on simulated data and apply it to data on extended-spectrum beta-lactamase (ESBL)-producing E. coli collected in Blantyre, Malawi. We compare the Rippler algorithm to two other commonly used Bayesian inference methods for partially observed epidemic data, and find that it has a good balance between mixing speed and computational complexity.

Related

• Real time analysis of epidemic data
• Inference on Extended-Spectrum Beta-Lactamase Escherichia coli and Klebsiella pneumoniae data through SMC²
• Stochastic Epidemic Models inference and diagnosis with Poisson Random Measure Data Augmentation
• Inference of Stochastic Disease Transmission Models Using Particle-MCMC and a Gradient Based Proposal
• Accelerating Bayesian inference for stochastic epidemic models using incidence data