Title: Fixed bandwidth CUSUM tests for change-in-mean under long memory
Authors: Christian Leschinski - Leibniz University Hannover (Germany)
Kai Wenger - Institute of Statistics (Germany) [presenting]
Abstract: Testing for mean-shifts in time series regression where the errors may exhibit long memory is considered. We propose four modified versions of the CUSUM test that apply kernel-based fixed-$b$ and fixed-$M$ long-run variance estimators. It is shown that the test statistics have a well-defined limiting distribution under long-range dependence that only depends on the long-memory parameter. We further discuss the bandwidth choice of all tests and show in an extensive Monte Carlo simulation study that our procedures perform best among all existing change-in-mean tests under long memory.