Log-Periodogram estimation of the memory parameter of a long-memory process under trend

Abstract

We show that small trends do not influence log-periodogram based estimators for the memory parameter in a stationary invertible long-memory process. In the case of slowly decaying trends which are easily confused with long-range dependence we show by Monte Carlo methods that the tapered periodogram is quite robust against these trends and thus provides a good alternative to standard log-periodogram methodology.