A stabilized bandwidth selection method for kernel smoothing of the periodogram

One popular method for nonparametric spectral density estimation is to perform kernel smoothing on the periodogram, and one important component of this method is the choice of the bandwidth (or span) for smoothing. This paper proposes a new bandwidth selection method that is based on a coupling of the so-called plug-in and the unbiased risk estimation ideas. This new method is easy to describe, simple to implement, and does not impose severe conditions on the unknown spectrum. Numerical results suggest that this new method often outperforms some other commonly used bandwidth selection methods. The new methodology is also applied to choose the bandwidth for log-periodogram smoothing. ( 2001 Elsevier Science B.V. All rights reserved.