Robin Eriksson
Department of Information Technology, Uppsala University
The use of data in support of decisions regarding public health is proving to become a necessity. One issue at hand is how we should prepare for epidemiological outbreaks. Currently, in our work, we study the spread of a verotoxigenic E. coli in the Swedish cattle population. We parameterize a disease-spread model by combining the high-performance simulator SimInf with actual agent transport and bacterial testing data. We perform Bayesian inference by using Approximate Bayesian Computations (ABC) and Synthetic Likelihood Adaptive Metropolis (SLAM) and obtain robust posterior parameter distribution. Finally, we show how one can leverage the obtained posterior and achieve credible intervals in evaluative experiments, e.g., intervention and detection methods.