The use of seroprevalence data to estimate cumulative incidence of infection

Abstract

This chapter introduces approximate Bayesian computation methods for likelihood free inference with particular emphasis on its application to epidemic models. A number of extensions of the basic approximate Bayesian computation algorithm are introduced to improve the performance of the algorithm. These extensions include how parameters are chosen for simulations, how simulations can be made more efficient, and bias correction techniques. The approximate Bayesian computation methodology and extensions are illustrated with two example datasets. The first dataset consists of the total number of cases in a measles outbreak in a Finnish school with a focus on estimating the transmission rate and efficacy of vaccination. The second dataset consists of two snapshots of a spatial epidemic in a citrus orchard with focus on the estimation of the transition kernel. Throughout both examples the simplicity of implementing approximate Bayesian computation is illustrated and the benefits of refining approximate Bayesian computation to make the algorithm more efficient are highlighted.