Dhorasso Junior Temfack Nguefack (Trinity College Dublin)
will speak on
Efficient sequential Bayesian inference for state-space epidemic models using ensemble data assimilation
Time: 3:00PM
Date: Thu 5th February 2026
Location: E0.32 (beside Pi restaurant)
[map]
Abstract: Estimating latent epidemic states and parameters from partial data is challenging. State-space models provide a probabilistic framework, but fully Bayesian estimation is computationally expensive because the likelihood requires integrating over all latent trajectories. Sequential Monte Carlo squared (SMC²) enables joint state-parameter inference by combining an outer SMC sampler with an inner particle filter, but the nested structure is computationally prohibitive for real-time outbreak response. We propose Ensemble SMC² (eSMC²), which replaces the inner particle filter with an Ensemble Kalman Filter (EnKF) to approximate the incremental likelihood. While this introduces bias via Gaussian approximation, we mitigate finite-sample effects using an unbiased Gaussian density estimator and adapt the EnKF for epidemic data through state-dependent observation variance. Simulations with known ground truth and application to 2022 U.S. monkeypox data show that eSMC² achieves substantial computational gains while producing posterior estimates comparable to SMC². The method offers an efficient framework for sequential Bayesian epidemic inference from imperfect surveillance data.
(This talk is part of the Working Group on Statistical Learning series.)
PDF notice
Return to all seminars