Xiahui Li (University of St Andrews)

Speaker: Xiahui Li (University of St Andrews)
Title: A Novel Approximate Bayesian Inference Method for Compartmental Models in Epidemiology using Stan
Abstract: Mechanistic compartmental models are foundational tools in epidemiology, offering critical insights into infectious disease dynamics and informing public health interventions during pandemics. Despite their utility, the growing complexity of these models and the increasing number of parameters required to capture rapidly evolving transmission scenarios present significant challenges for parameter estimation, particularly when likelihoods are intractable. Likelihood-free methods, such as Approximate Bayesian Computation (ABC) and Bayesian Synthetic Likelihood (BSL), have emerged as effective approaches for addressing these challenges. In this study, we propose a novel hybrid framework that integrates the strengths of ABC, BSL, and Hamiltonian Monte Carlo (HMC) to enhance parameter inference for complex epidemiological models. The method employs ABC to systematically select an informative subset of summary statistics, which are then used to construct a synthetic likelihood under the BSL framework. HMC, implemented via the Stan software, is utilized for efficient posterior sampling, leveraging gradient-based techniques to explore high-dimensional parameter spaces. The proposed framework is validated through numerical studies, demonstrating its robustness, accuracy, and computational efficiency.