Block Bootstrapping the Empirical Distance Covariance
share
We prove the validity of a non-overlapping block bootstrap for the empirical distance covariance under the assumption of strictly stationary and absolutely regular sample data. From this, we develop a test for independence of two strictly stationary and absolutely regular processes. In proving our results, we derive explicit bounds on the expected Wasserstein distance between an empirical measure and its limit for strictly stationary and strongly mixing sample sequences.
READ FULL TEXT