Event Details

A mathematics talk for a general audience, given by a member of the School of Mathematical Sciences.

Speaker: Lloyd Chapman

Title: Fitting epidemic models to data: why it matters and why we could still be better at it

Abstract: Infectious disease models play an essential role in predicting the course of an outbreak and how different interventions may affect it, particularly during an epidemic or pandemic. To do this reliably, they need to be fitted to data. However, the disease transmission process is only partially and noisily observed, e.g. we do not observe when individuals get infected, only when they develop symptoms (if they do) or test positive, and in an ongoing outbreak the data is incomplete. Furthermore, policymakers often want answers to questions that require fitting detailed or complex models to the observed data. Together, these make fitting infectious disease models very challenging, as the algorithms used have to be able to efficiently explore a very high dimensional space of model parameters and missing data and handle the strong correlation that exists between the two.

In this talk, I will give an overview of the history of Bayesian approaches to fitting infectious disease models to data and their application to different epidemics and diseases. I will talk about the approaches that have been taken to deal with the above challenges, how they don't quite fully solve the problem and ongoing work to try address the gap.