Prediction and tracking of long-range-dependent sequences

Long-range-dependent (LRD) sequences have been found to be of importance in various fields such as telecommunications, signal processing and finance. Since the history of an LRD sequence has significant impact on the present values, it is expected that accurate prediction and tracking of these sequences are easier than of short-range-dependent sequences. The purpose of this paper is to verify whether distant observations in the past might increase the performance of a constrained tracker significantly when this information from the past is used in combination with recent observations.