Abstract

Zoom link: https://maths-cam-ac-uk.zoom.us/j/94018037756

I will present a scheme through which stochastic compartmented models for the spread of an epidemic can be specified and calibrated in a fully Bayesian fashion using surveillance data. In order to account for imperfections in the reported data, we integrate in our models the process by which individuals get tested, which is informed by available data on the number of tests performed. For the likelihood estimation, we take into account all sources of stochasticity in both the infection and the testing process, as well as resulting correlations in the data. I will illustrate this scheme using the latest data for the covid19 pandemic in France, Germany, and the UK, allowing us to discuss the effect of non-pharmaceutical interventions. As an outlook, I will show how vaccinations can be modeled as a finite resource similar to testing.