B0215
Title: Multivariate spurious long memory and a robust local Whittle estimator
Authors: Philipp Sibbertsen - University of Hannover (Germany)
Christian Leschinski - Leibniz University Hannover (Germany) [presenting]
Abstract: For univariate time series it is well documented that low frequency contaminations generate spurious long memory. This analysis is extended to vector valued processes. A multivariate generalization of the random level shift process is introduced and its properties are derived. These results show, that spurious long memory is an issue for vector series as it is for univariate series. Therefore, a robust multivariate local Whittle (RMLW) estimator is derived that is robust to low frequency contaminations and asymptotically normal. A Monte Carlo study shows that the RMLW estimator has good finite sample properties. The standard Gaussian semiparametric estimator, on the other hand, is severely biased.
Journal