A New Estimator of the Spectral Density Function

Abstract

In the study of stationary processes on the real line, the spectral density function is a parameter of considerable interest. In this talk, we discuss a new estimator of the spectral density function obtained by an algorithmic inversion of estimated covariances. In particular, the data are not required to be observed on a grid and the estimator is not based on the periodogram. For data that are observed on a grid, the estimator can be derived in closed from, and the mean squared error of the estimator can be computed.