B1215
Title: Testing independence of mixing time series using the distance covariance
Authors: Annika Betken - University of Twente (Netherlands)
Herold Dehling - Ruhr-University Bochum (Germany)
Marius Kroll - Ruhr-University Bochum (Germany) [presenting]
Abstract: A test for independence of absolutely regular time series is developed based on an independent block bootstrap for the distance covariance. The resulting test can detect any dependence up to a pre-specified lag and outperforms other tests of independence, e.g. those based on Pearson's correlation. New bounds are proved on the Wasserstein distance between the empirical measure of a strongly mixing stationary process and its marginal distribution and are readily generalized to bootstrap procedures of different V-statistics. The bootstrap procedure may be adapted to yield confidence intervals for the distance covariance. Simulations suggest that it performs better than approaches based on the sample variance of the point estimator for the distance covariance.
