This talk has been automatically captioned. You can remove these by pressing CC on the video toolbar.

Name: Gloria Buriticá

Talk Title: Modeling clusters of extreme events over short periods

Abstract: We study the extreme dependence structure of stationary regularly varying time series. The classical approach for modeling extremes in this setting is based on the identification of short periods with several exceedances also known as clusters of extremes. The main summary of the clustering phenomenon is given by the extremal index. We suggest studying a general version of clustering where we consider short periods for which a modulus function applied to the vector of consecutive days exceeds a high threshold. Then for a convenient choice of modulus, the estimation of the extremal index can be avoided. We deduce an inference method for computing extreme returns levels that avoids the computation of the extremal index. We show applications for precipitation data in France.

This talk is a contributed talk at EVA 2021.