Bayesian inference of model discrepancy in epidemiological systems
Résumé
Lack of data and information for parameters is a serious problem for epidemiological applications. The use of probabilistic models allows analyse the uncertainties induced by this lack of knowledge in the modeling process. This work is applies a methodology to deal with the model errors in a epidemiological system employing a Polynomial Chaos Expansion to represent model discrepancy and Bayesian inference to learn its coefficients. Maximum Entropy Principle constructs prior distribution and the effects of several Likelihoods are compared.