An Estimating Method for Parametric Spectral Densities of Gaussian Time Series
A Gaussian time series \(x_N=\{x(1),\dots,x(N)\}\) with zero mean and spectral density \(S(\lambda,\theta)\) is considered, where \(\theta\in\Theta\subseteq R^p\) is an unknown parameter. The problem of asymptotically efficient \(S(\lambda,\theta)\) estimation is considered with the Kullback-Leibler divergence as the loss function. It is common to use as an estimator \(S(\lambda,\hat\theta)\) where \(\hat\theta\) is an asymptotically efficient estimator of \(\theta\). There exist estimators of the form \[ \bar S_\Delta(\lambda,\hat\theta)=S(\lambda,\hat\theta)+N^{-1}\Delta(\lambda,\hat\theta)+o_p(N^{-1}), \] which are asymptotically better then \(S(\lambda,\hat\theta)\) for any asymptotically efficient estimator. Here \(\Delta\) is the mixture mean curvature vector in the differential geometry of the estimation problem. Note that \(\bar S_\Delta\) is not of the form \(S(\lambda,\theta)\). The author calls these estimators predictor estimators. For the case of AR(p) processes \(x_N\) the author describes the predictor estimators which are the spectral densities of ARMA processes with AR-order \(3p\). The author discusses also the \(\alpha\)-divergence as a loss function for the spectral density estimators.
- Optimal estimation of power spectral density by means of a time-varying autoregressive ap\-proach.
- Superharmonic priors for autoregressive models
- Asymptotic expansion of the risk difference of the Bayesian spectral density in the autoregressive moving average model
- A superharmonic prior for the autoregressive process of the second-order
- scientific article; zbMATH DE number 3896148 (Why is no real title available?)
- scientific article; zbMATH DE number 1423407 (Why is no real title available?)
This page was built for publication: An Estimating Method for Parametric Spectral Densities of Gaussian Time Series
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4258764)
