Uncertainty quantification for epidemiological forecasts of COVID-19 through combinations of model predictions
share
A common statistical problem is prediction, or forecasting, in the presence of an ensemble of multiple candidate models. For example, multiple candidate models may be available to predict case numbers in a disease epidemic, resulting from different modelling approaches (e.g. mechanistic or empirical) or differing assumptions about spatial or age mixing. Alternative models capture genuine uncertainty in scientific understanding of disease dynamics, and/or different simplifying assumptions underpinning each model derivation. While the analysis of multi-model ensembles can be computationally challenging, accounting for this 'structural uncertainty' can improve forecast accuracy and reduce the risk of over-estimated confidence. In this paper we look at combining epidemiological forecasts for COVID-19 daily deaths, hospital admissions, and hospital and ICU occupancy, in order to improve the predictive accuracy of the short term forecasts. We combining models via combinations of individual predictive densities with weights chosen via application of predictive scoring, as commonly applied in meteorological and economic forecasting.
READ FULL TEXT
