Broad band semiparametric estimation of the memory parameter of a long-memory time series using fractional exponential models (Q2740106)

The authors consider a Gaussian long-memory time series with spectral density NEWLINE\[NEWLINEf(\lambda)=|1-e^{-i\lambda }|^{-2d}f^*(\lambda),\quad \lambda \in [-\pi;\pi],NEWLINE\]NEWLINE where the memory parameter \(d\in (-1/2, 1/2)\) and the function \(f^*(\lambda)\) govern the long- and short-term correlation structures of the series, respectively. A semiparametric estimator of \(d\) is considered. The bias and the variance of the estimator are derived. Some of their asymptotic properties are given. A Monte Carlo study is presented.