On the Selection of an Optimum Spectral Window for Analysis of Certain Classes of Time Series Data

摘要

The study is aimed at providing insight into selection of optimum spectral windows appropriate to different populations representing oceanographic and geophysical events. From solution of 2nd order and 4th order differential equations, two single-peak and two double-peak spectra are derived. Locations of peak frequency and bandwidth are varied to provide four different representative populations. For each, time histories are generated, and spectral estimates are made for different lengths of record and for different resolution, with 6 spectral (lag) windows applied. It is found that for total variance, any of the spectral windows would give a good estimate. Where shape of the spectrum is important, the DTMB and Prazen (II) windows tend to provide the best approximation to the true spectrum. The integrated mean square error is used as the figure of merit in evaluating the efficacy of the different windows. (Author)