Study on Order of Autoregressive Model for Spectral Analysis of Long Period Magnetic Pulsations

概要

論文の詳細を見る
The autoregressive (AR) model has been applied to spectral analysis for various geophysical data, since it has a high frequency resolution especially for short length of the data. In order to get superiority with the high frequency resolution, it is very important to select the order of the AR process. Although a few criteria for determining the order have been proposed, they are not perfect for spectral analysis of long period pulsations. In this paper we clarify the property of the AR spectrum by using test data, which are simulated as a wave form of magnetic pulsations having two frequency components. We calculate AR spectra for the test data by changing the order of tht AR process, and examine the relationship between the frequency resolution of spectral peak and the order of the AR process. It is found that the appropriate order of AR is strongly correlated with the dominant frequency included in the signal to be analyzed. We propose a new criterion to estimate the appropriate order of AR in the frequency range of pulsations being analyzed.
東海大学の論文