CMStatistics 2022: Start Registration
View Submission - CMStatistics
Title: Extreme value modelling of SARS-CoV-2 community transmission using discrete Generalised Pareto distributions Authors:  Abdelaati Daouia - Fondation Jean-Jacques Laffont (France) [presenting]
Gilles Stupfler - University of Angers (France)
Antoine Usseglio-Carleve - Avignon Université (France)
Abstract: Superspreading has been suggested to be a major driver of overall transmission in the case of SARS-CoV-2. It is, therefore, important to statistically investigate the tail features of superspreading events (SSEs) to understand virus propagation and control better. Our extreme value analysis of different sources of secondary case data indicates that SSEs associated with SARS-CoV-2 may be fat-tailed, although substantially less so than predicted recently in the literature, but also less important relative to SSEs associated with SARS-CoV. The results caution against pooling data from both coronaviruses. This could provide policy- and decision-makers with a more reliable assessment of the tail exposure to SARS-CoV-2 contamination. Going further, we consider the broader problem of large community transmission. We study the tail behaviour of SARS-CoV-2 cluster cases documented both in official reports and in the media. Our results suggest that the observed cluster sizes have been fat-tailed in the vast majority of surveyed countries. We also give estimates and confidence intervals of the extreme potential risk for those countries. A key component of our methodology is up-to-date discrete Generalised Pareto models, which allow for maximum-likelihood-based inference of data with a high degree of discreteness.