Count regression models for COVID-19

Stephen Chan, Jeffrey Chu, Yuanyuan Zhang, Saraleesan Nadarajah

Research output: Contribution to journalArticlepeer-review

Abstract

At the end of 2019, the current novel coronavirus emerged as a severe acute respiratory disease that has now become a worldwide pandemic. Future generations will look back on this difficult period and see how our society as a whole united and rose to this challenge. Many reports have suggested that this new virus is becoming comparable to the Spanish flu pandemic of 1918. We provide a statistical study on the modelling and analysis of the daily incidence of COVID-19 in eighteen countries around the world. In particular, we investigate whether it is possible to fit count regression models to the number of daily new cases of COVID-19 in various countries and make short term predictions of these numbers. The results suggest that the biggest advantage of these methods is that they are simplistic and straightforward allowing us to obtain preliminary results and an overall picture of the trends in the daily confirmed cases of COVID-19 around the world. The best fitting count regression model for modelling the number of new daily COVID-19 cases of all countries analysed was shown to be a negative binomial distribution with log link function. Whilst the results cannot solely be used to determine and influence policy decisions, they provide an alternative to more specialised epidemiological models and can help to support or contradict results obtained from other analysis.
Original languageEnglish
Article number125460
Pages (from-to)1-10
Number of pages10
JournalPhysica A: Statistical Mechanics and its Applications
Volume563
Early online date31 Oct 2020
DOIs
Publication statusPublished - 1 Feb 2021

Keywords

  • coronavirus
  • epidemiology
  • negative binomial distribution
  • Poisson distribution

Fingerprint

Dive into the research topics of 'Count regression models for COVID-19'. Together they form a unique fingerprint.

Cite this