The Fast Fourier Transform for Experimentalists Part IV: Autoregressive Spectral Analysis

Published

November 30, 2005

Author(s)

Bert W. Rust, D Donnelly

Abstract

This tutorial paper is the fourth in a series devoted to the use of the Fast Fourier Transform (FFT) in time series analysis. It describes the parametric methods for estimating the power spectral density (PSD) that are used when the time series is assumed to be well modelled by an autoregressive process. In such cases, the PSD estimates can be calculated from estimates of the autoregressive parameters. A special case is the Maximum Entropy Method (MEM) which seeks the parameter estimates which minimize the assumptions about the data outside the window of observation. In all of these methods, the results are strongly dependent on the choice of the order of the autoregressive process. Two simple noisy time series are use to illustrate these issues.

Citation

Computing in Science & Engineering

Volume

7

Issue

6

Pub Type

Journals

Download Paper

Local Download

Keywords

autoregressive spectral analysis, fast fourier transform, maximum entropy method, power spectral density

Citation

Rust, B. and Donnelly, D. (2005), The Fast Fourier Transform for Experimentalists Part IV: Autoregressive Spectral Analysis, Computing in Science & Engineering, [online], https://tsapps.nist.gov/publication/get_pdf.cfm?pub_id=51316 (Accessed January 21, 2025)