B1003
Title: Calibrating a large-scale stochastic meta-population model of COVID-19 in England and Wales
Authors: Trevelyan McKinley - University of Exeter (United Kingdom) [presenting]
Abstract: Calibration of complex stochastic infectious disease models is challenging. These often have high-dimensional input spaces, with the models exhibiting complex, non-linear dynamics. Coupled with this is a paucity of necessary data, resulting in a large number of hidden states. Methods based on simulating the hidden states directly from the model-of-interest have the advantage that they are often easier code than likelihood-based approaches, and thus models can be developed and adapted quickly. However, they can be extremely computationally intensive, often requiring very large numbers of simulations in order to adequately explore the input space; rendering them infeasible for many large-scale problems. We extend recent developments in emulation-based calibration methods to calibrate a large-scale, stochastic, age-structured meta-population model of COVID-19 transmission in England and Wales. This approach, called history matching, develops quick surrogate models (emulators) that can be used in place of the complex simulator to efficiently explore the input space, and remove parts of the space where model fits are unlikely to be found. It does this whilst accounting for important sources of uncertainty: such as observation error, simulator uncertainty, emulator uncertainty and model discrepancy. We discuss various challenges relating to the model in question, and discuss the feasibility of implementing these methods for future pandemic modelling efforts.
