POT-flavored estimator of Pickands dependence function

TL;DR

This paper introduces a novel estimator combining Peak-Over-Threshold and Block-Maxima methods to improve the estimation of the Pickands dependence function in bivariate time series, reducing bias and mean squared error.

Contribution

It presents a new estimator that integrates two flavors of extreme value analysis to enhance accuracy in dependence function estimation.

Findings

Reduces asymptotic bias in estimation

Lowers overall mean squared error

Effective for bivariate time series

Abstract

This work proposes an estimator with both Peak-Over-Threshold and Block-Maxima flavors, uses it to estimate the Pickands dependence function of bivariate time series, and illustrates how it brings down the asymptotic bias and the overall mean squared error.