The Stationary Autogressive Model
Abstract
There are several ways of developing the autoregressive stochastic process as a finite-parameter model for time series analysis. This paper obtains the properties of the autoregressive process from a stationary stochastic process that satisfies the simple condition that a linear combination of current and past elements of the process is independent of (or alternatively uncorrelated with) all earlier elements of the process. This approach provides a coherent, clear, and rigorous exposition of the autoregressive model. The stationarity and independence imply that the roots of the associated polynomial equation are less than 1 in absolute value. The existence of the moving average representation is deduced and its form for distinct roots. The Yule-Walker equations, which are derived, determine the autocovariance sequence. Another set of parameters consists of the variance of the process and the partial autocorrelation sequence.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jul 01, 1986
- Accession Number
- ADA171416
Entities
People
- Theodore W. Anderson
Organizations
- Stanford University