An Exponential Model for the Spectrum of a Scalar Time Series
Abstract
A new class of parametric models for the spectrum of a scalar time series is proposed, in which the logarithm of the spectral density function is represented by a finite Fourier series. Two alternative parameter estimation procedures are described, and the use of a fitted model to provide forecasts of future values is discussed. The model has been compared with the more conventional autoregressive/moving-average model, and the results of their comparison are given.
Document Details
- Document Type
- Technical Report
- Publication Date
- Apr 01, 1972
- Accession Number
- AD0746160
Entities
People
- P. Bloomfield
Organizations
- Princeton University