On the Estimation of the First-Order Autoregressive Parameter.

Abstract

This paper presents Cramer-Rao (C-R) bounds for parameter estimation of a first-order autoregressive process from a finite record of data. The authors used these bounds to evaluate the performance of Maximum Likelihood Estimation (MLE) and linear prediction approaches. Some estimators use low-rank approximation of an estimated covariance matrix. The latter estimates are based on the method of Tufts and Kumaresan. In this document a zero selection techniques in the last step of the procedure was added. The low-rank, high order, linear prediction estimator performs better than the other methods which were tested, when the pole is close to the unit circle. It is slightly biased and its variance is small and close to the variance given by the C-R bound for unbiased estimators. For a small number of samples (25 to 100) this estimator performs substantially better than the MLE. (Author)

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Jun 01, 1984
Accession Number
ADA144284

Entities

People

  • D. W. Tufts
  • F. Giannella

Organizations

  • University of Rhode Island

Tags

Communities of Interest

  • Materials and Manufacturing Processes

DTIC Thesaurus Topics

  • Coefficients
  • Covariance
  • Eigenvalues
  • Eigenvectors
  • Electrical Engineering
  • Engineering
  • Errors
  • Estimators
  • Maximum Likelihood Estimation
  • Military Research
  • Noise
  • Observation
  • Polynomials
  • Rhode Island
  • Simulations
  • Universities
  • White Noise

Fields of Study

  • Engineering

Readers

  • Linear Algebra
  • Statistical inference.