An Effective ARMA Modeling Method.
Abstract
The ability to model random time series plays a prominent role in a variety of applications as exemplified by seismic data analysis, doppler radar processing, speech processing, adaptive filtering, and, array processing. Undoubtedly, two of the more popular procedures for effecting such time series models are the classical Fourier (MA) approach and the maximum entropy (AR) method. In this paper, a theoretical comparison of these contemporary procedures with a more general ARMA method will be made. It will be demonstrated that the spectral estimation performance of the ARMA method typically exceeds that of its more special-method typically exceeds that of its more specialized MA and AR counterparts. With this supremacy thus established, a recently developed method for effectively generating ARMA model estimates from time series observations will be then presented. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Apr 01, 1981
- Accession Number
- ADA101871
Entities
People
- James A. Cadzow
Organizations
- Virginia Tech