Extreme-value modelling of daily reported COVID-19 incidence in Zambia: tail behaviour, clustering, and return levels
Discover Informatics · 20.11.2026 · 1 (1)
DOI: 10.1007/s44564-026-00046-7 ↗Kurz gefasst
Fachartikel in der Zeitschrift Discover Informatics, online erschienen am 20.11.2026. Verfasst von Aaron Chola und 5 weiteren Autor:innen. Die Arbeit nennt 30 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
The COVID-19 burden was shaped not only by average transmission levels, but also by short periods of unusually high incidence that placed severe pressure on health systems and public health services. In this study, Extreme Value Theory was applied to quantify the magnitude, frequency, and clustering of extreme daily COVID-19 incidence (new confirmed cases per million population) in Zambia from 4 January 2020 to 31 December 2023. We combined two complementary approaches for our EVT analysis: the block maxima method using the Generalised Extreme Value distribution, and the peaks-over-threshold method fitted with the Generalised Pareto distribution. For the block maxima analysis, block lengths of 7, 14, 21, and 28 days were assessed, with 48 positive block maxima obtained for the selected 28-day block. The 28-day block was selected based on combined diagnostic evidence, including serial-dependence and goodness-of-fit assessments. The fitted GEV model produced a positive shape parameter, $$\xi =0.51$$ (95% CI 0.29 to 0.61). For the peaks-over-threshold analysis, the 90th percentile threshold, $$u=61.26$$ , was selected using mean residual life plots and parameter stability assessment. This produced 83 raw threshold exceedances, which were reduced to five independent clusters after declustering. The estimated extremal index $$\theta =0.06$$ (95% CI 0.04 to 0.22), indicating strong clustering. The fitted GPD produced a shape parameter of $$\xi = 0.56$$ (95% CI −2.43 to 0.82), indicating substantial uncertainty in the estimated tail shape; consequently, the GPD analysis was treated as exploratory rather than as primary evidence of heavy-tailed behaviour. The 2- and 5-year GEV return levels were 149.69 and 254.55 cases per million, respectively, while the corresponding GPD estimates were 100.72 and 166.81 cases per million. Longer-horizon return level estimates involve substantial extrapolation and should therefore be interpreted with caution. These Extreme Value estimates provide useful metrics that may help with epidemic preparedness, hospital surge capacity planning, and resource allocation during public health emergencies.
Zitieren
APA
Chola, A., Hamweene, O., Kabaghe, P., Chisanga, P. M., Moyo, E., & Babasola, O. (2026). Extreme-value modelling of daily reported COVID-19 incidence in Zambia: tail behaviour, clustering, and return levels. Discover Informatics, 1(1). https://doi.org/10.1007/s44564-026-00046-7
Vancouver
Chola A, Hamweene O, Kabaghe P, Chisanga PM, Moyo E, Babasola O. Extreme-value modelling of daily reported COVID-19 incidence in Zambia: tail behaviour, clustering, and return levels. Discover Informatics. 2026;1(1). doi:10.1007/s44564-026-00046-7
BibTeX
@article{Chola2026,
author = {Chola, Aaron and Hamweene, Obvious and Kabaghe, Paul and Chisanga, Patience Mulenga and Moyo, Edwin and Babasola, Oluwatosin},
title = {Extreme-value modelling of daily reported COVID-19 incidence in Zambia: tail behaviour, clustering, and return levels},
journal = {Discover Informatics},
year = {2026},
volume = {1},
number = {1},
doi = {10.1007/s44564-026-00046-7}
}
EUVIMED