CMStatistics 2015: Start Registration
View Submission - CMStatistics
B1387
Title: Weighted tests for distributional change in long-memory processes Authors:  Johannes Tewes - Ruhr-Universitaet Bochum (Germany) [presenting]
Abstract: We study changes in the marginal distribution of a subordinated Gaussian process that exhibits long-range dependence. Tests for such change-points are occasionally based on Kolmogorov-Smirnov or Cramer-von Mises statistics. However, our simulation study indicates that the combination of strong dependence and change-points occurring near the beginning or near the end of the observation period drastically reduces the power of these tests. We therefore consider weighted versions of the statistics. In order to calculate critical values we show that the weighted sequential empirical process converges in distribution to a semi-degenerate process. The structure of the limiting process, e.g. the marginal distribution and the covariance structure, mainly depends on the Hemite rank. It is the index of the first nonzero coefficient in the Hermite expansion of one summand of the empirical process. A special feature of distributional changes is the fact that the Hermite rank may change, too. We consider local alternatives covering this scenario, and as a result, we may derive the asymptotic power of the change-point tests.