CMStatistics 2021: Start Registration
View Submission - CFE
A0865
Title: Nonparametric estimator of tail dependence coefficients: Balancing bias and variance Authors:  Maxime Nicolas - Universite Paris 1 Pantheon-Sorbonne (France) [presenting]
Matthieu Garcin - Leonard de Vinci Pole Universitaire (France)
Abstract: In risk management, the Tail Dependence Coefficient (TDC) is generally used to measure the dependence between extreme events. Recent work has focused on the non-parametric estimators of the TDC, depending on a threshold that defines which rank corresponds to the tails of a distribution. We present a new method to select the threshold according to a tradeoff between the bias and variance of the estimator. It combines the theoretical Mean Squared Error (MSE) of the estimator and a parametric estimation of the copula linking observations in the tails. We develop several estimation procedures and compare them with other common estimators in a simulation framework. Finally, the method is used to provide risk measurement in a financial dataset.