Lower Bound on the Risk for Spectral Density Estimates

The problem of estimating the spectral density of a stationary Gaussian random sequence is considered. A minimax lower bound on the risk for the spectral density estimates is obtained that is exact in order of magnitude as the number of observations increases.