Discussion paper series // Zentrum für Finanzen und Ökonometrie, Universität Konstanz 2008,10
Filtered log-periodogram regression estimation of the fractional differencing parameter d is considered. Asymptotic properties are derived and the effect of filtering on d is investigated. It is shown that the estimator by Geweke and Porter-Hudak (1983) can be improved significantly using a simple family of filters. The essential improvement is based on a binary decision that is asymptotically correct with probability one. The idea is closely related to the well known technique of pre-whitening.
Filtering log-periodogram regression local pre-whitening fractional differencing parameter long memory