Recently, we have proposed a general two-pathogen single-host model to examine the dynamical consequences of alternative modes of pathogen interactions. The framework allows a mechanistic exploration of interference and facilitation effects of pathogens mediated via ecological and/or immunological processes. In this talk we present a computational statistics framework that allows the inference of interaction parameters from time series of infectious diseases using our general model. Emphasis in our talk will be on the use of likelihood and Bayesian methods utilizing Monte Carlo Markov Chain methods and efficient integration of large stochastic epidemiological models.
The method allows taking advantage of readily available statistical summaries of time series data to efficiently integrate complex likelihoods. Our method is flexible enough to apply to a broad range of stochastic population and epidemiological models. It allows simultaneous estimation of multiple parameters for models with many state variables and parameters. Understanding the dynamic implications of multi-pathogen associations has potentially important public heath consequences: integration of such a framework within an efficient statistical computing platform, combined with a strong data base is one path towards a predictive theory of multi-pathogen epidemics.
