The bootstrap applied to power spectral density function estimation

Abstract

A shortcoming in the use of autoregressive spectral estimation has been the lack of methods for placing confidence intervals or bands on estimated spectral density. The bootstrap method is applied to obtain nonparametric confidence bands and intervals. It is the aim of this paper to indicate how this can be done and to report on the outcome of some Monte Carlo studies of the properties of the resulting procedures. Some asymptotic theory concerning the properties of the bootstrap confidence band is also discussed.