Forecasting and Whitening Filter Estimation.
Abstract
An approach to empirical time series analysis is described in which the identification stage is not accomplished chiefly by graphical inspection of the time series and of computed auxiliary sample functions such as the autocorrelation function, partial autocorrelation function, and spectrum. Rather the transfer function g sub(infinity) of the whitening filter is directly estimated and parsimoniously parametrized. A criterion for choosing a regression model for forecasting is described. A model identification procedure for a stationary time series is described. Model identification for a non-stationary time series is discussed. Our approach is illustrated by an example.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jun 01, 1977
- Accession Number
- ADA056540
Entities
People
- Emanuel Parzen
Organizations
- University at Buffalo