Scientific Computation and Applied & Industrial Mathematics: Mike Irvine

Complex individual-based models abound in epidemiology and ecology. Fitting these models to data is a challenging problem: methodologies can be inaccessible to all but specialists, there may be challenges in adequately describing uncertainty in model fitting, and the complex models may take a long time to run, requiring parameter selection procedures. Approximate Bayesian Computation has been proposed as a likelihood-free method in resolving these issues, however requires careful selection of summary statistics and annealing scheme. I compare this procedure directly to standard methodologies where the likelihood exists, Markov-chain Monte Carlo and maximum likelihood. This is then applied to a complex individual-based simulation for lymphatic filariasis, a human parasitic disease, which affects over 120 million individuals internationally. Finally, I will discuss a new approach to individual-based model fitting by constructing a synthetic likelihood using mixture density networks.