Log-average periodogram estimator of the memory parameter

TL;DR

This paper proposes a semiparametric regression estimator for the long-memory parameter in time series, using periodogram averaging near zero frequency, with theoretical backing and promising empirical results.

Contribution

It introduces a novel semiparametric estimator based on periodogram averaging, providing theoretical justification and empirical evidence of effectiveness.

Findings

Estimator is theoretically justified.

Monte Carlo simulations show promising accuracy.

Method outperforms some existing estimators.

Abstract

This paper introduces a semiparametric regression estimator of the memory parameter for long-memory time series process. It is based on the regression in a neighborhood of the zero-frequency of the periodogram averaged over epochs. The proposed estimator is theoretically justified and empirical Monte Carlo investigation gives evidence that the method is very promising to estimate the long-memory parameter.