Abstract
The purpose of the present paper is to introduce a method for forecasting daily and total numbers of COVID-19-associated deaths. We apply the Gumbel distribution function for the analysis of time series data of the first wave. The Gumbel distribution function F(t) has a notable property of F(t) = 1/2.718 at the node (peak) point of the distribution. Therefore, we can forecast the number of total deaths N. In the present study, the Gumbel model gives the estimate N ≈ 2.718N1, where N1 is the partial sum of the daily numbers of deaths until the day of the peak. The proposed model can also forecast the daily numbers after the peak. We investigated the data of New York City, Belgium, Switzerland, Sweden, and the United Kingdom. The Gumbel model gives reasonable results for New York City, Belgium, and Switzerland. On the other hand, the proposed method underestimates N for Sweden and the United Kingdom. The proposed approach is very simple, and carrying out the analysis is easy. This method uses spreadsheet software for most of the calculations, and no special program is needed.
An Analysis of Record Statistics based on an Exponentiated Gumbel Model