Uncertainty Quantification for Expensive COVID-19 Simulation Models (UQ4Covid)

The UQ4Covid project develops uncertainty quantification (UQ) techniques for models of COVID-19 in the UK. The project brings together epidemiologists and statisticians to use high-resolution computer models for policy-relevant decisions.

Advanced UQ tools allow us to

• rapidly re-calibrate computer models to respond to new data and changes in the evolution of the pandemic;
• efficiently evaluate different interventions such as temporary localised lockdowns, vaccination strategies; and
• accurately represent prediction uncertainties due to incomplete data and imperfect simulation models.

UQ4Covid is an open science project. We use publicly available epidemiological models and statistical software packages. Findings are shared with the community in real time through a constantly updated list of vignettes, contributed by researchers from different UK universities. These vignettes document all analyses in detail and provide all source code necessary to make the results reproducible.

• Link to vignettes

Tools

• MetaWards: National-scale stochastic metapopulation model of disease transmission in Great Britain
• mogp: The Multi-Output Gaussian Process Emulator
• ExeterUQ_MOGP: An R interface to mogp.
• R-INLA: Large-scale spatio-temporal Bayesian Inference

COVID-19 Uncertainty Quantification using the MetaWards Model

COVID-19 Variant Classification