Covid-19 projections for Spain using forecast combinations

• Vilar Fernández, J.A. ^[1] ; Casal, R.F. ^[1] ; Fernandez-Lozano, C. ^[1]
  1. [1] Universidade da Coruña

     Universidade da Coruña

     A Coruña, España

     
• Localización: BEIO, Boletín de Estadística e Investigación Operativa, ISSN 1889-3805, Vol. 36, Nº. 2, 2020, págs. 99-125
• Idioma: inglés
• Enlaces
  • Texto Completo Ejemplar
• Resumen

  • As part of the “Mathematics against coronavirus” initiative promoted by the Spanish Committee for Mathematics (CEMat), an interactive web application based on R was developed to monitor and predict the short-term behavior of relevant variables in Covid-19 spreading. For every Spanish administrative region, predictions from a variety of models and techniques provided by independent research groups were combined to generate cooperative predictions, which have been daily available on the web together with the official data from the Institute of Health Carlos III (ISCIII). Since forecast combination can improve forecasting accuracy, particularly in contexts with high uncertainty, the goal was to use the forecasts of the Spanish mathematical community to obtain more accurate and stable predictions and, eventually, report conclusions to the authorities. This article provides a general overview of the development and results of this project, motivating the use of combined forecasts and including information on the main stages of the process. © 2020 SEIO