サロゲート法おける窓関数の効果(非線形回路とシステム,及び一般)

概要

論文の詳細を見る
In the case of making surrogate data in non-linear time series analysis, we usually use the Fourier transform. In the calculation, it is assumed that the time series is periodic. Since such an assumption does not always hold, estimation accuracy of the Fourier transform, then power spectra, becomes worse. Due to such estimation error, it is possible that surrogate test leads a false conclusion, for example, linear time series shows non-linear property. In the present paper, we experimentally evaluate effect of applications of data windows for better estimation of power spectra in the case of producing Fourier transformed, Fourier shuffled and iterative amplitude adjusted Fourier transformed surrogates. In order to examine the effect of data windows on false rejections, we perform significance testings, in which we consider autoregressive model as original data. Our results show that data windows make false rejection rates more reduced as the data length is shorter.
社団法人電子情報通信学会の論文
2003-09-23