A goodness-of-fit test for heavy tailed distributions with unknown parameters and its application to simulated precipitation extremes in the Euro-Mediterranean region

Abstract

We establish a general bootstrap procedure combined with a modified Anderson–Darling statistic. This procedure is proved to be valid for heavy tailed generalized Pareto distributions that are commonly used to model excesses over a high threshold in extreme value theory. Then, the method is applied to daily precipitation excesses simulated over the Euro-Mediterranean region in autumn by four regional climate models from the EURO-CORDEX initiative.