This talk has captions. You can remove these by pressing CC on the video toolbar.  

Name: Jochem Oorschot

Talk Title: Extreme U-statistics

Abstract: Extreme U-statistics (joint work with Johan Segers & Chen Zhou)

One-sample U-statistics generalize the notion of averages and lead to minimum-variance unbiased estimators. Segers (2001) considered U- for sequences of location-scale invariant kernels whose degree grows with the sample size and which depend only on the largest of their arguments. By ensuring that the kernel’s expectation converges to a function of the extreme value index, it is possible to construct consistent estimators of the latter using such Extreme U-statistics. The all-block maxima in Oorschot & Zhou follow a similar logic, by recording the maxima of all sub-samples of observations of a given size. In this paper we explore the asymptotic normality of  general Extreme-U-statistics. The theory draws largely from the one of U-statistics by considering kernels applied to blocks of intermediate length m(n) that grow as a function of the sample size n.

This talk is a contributed talk at EVA 2021. View the programme here.