Model fitting, inference and methods

Members of the CMMID have an interest in developing new methodology to underpin the complex models we use. This includes both methods of gathering data (e.g. Respondent-Driven Sampling) and approaches to model fitting, parametric inference and uncertainty and sensitivity analyses.

We have been involved in early work on the application of Bayesian Emulation to epidemic modelling. Emulation allows us to develop cheap surrogates for complex models, thus reducing the computational burden of traditional uncertainty and sensitivity analyses. Another interest is in the development and application of Approximate Bayesian Computation and particle-based Markov Chain Monte Carlo methods. These can be used to efficiently characterise the likelihood surface for models for which it is not possible to write the likelihood explicitly, allowing us to infer the true underlying distribution of model parameters. We are currently looking at ways to combine these two families of Bayesian methods in novel model-fitting techniques.


Ioannis Andrianakis, Marc Baguelin, Anton Camacho, Pete Dodd, John Edmunds, Sebastian Funk, Helen Johnson, Richard White


Short course on Model fitting and inference for infectious disease dynamics.


Calibration and analysis of complex individual-based stochastic models

Comments are closed.