Time Series Model Identification and Prediction Variance Horizon.
Abstract
An approach to time series modelling is described; it classifies the time into one of three memory types (called no memory, short memory, and long memory), and then finds a whitening filter. When the time series is short memory one would like to identify the whitening filter type as AR, MA, or ARMA before parameter estimation. A new tool is introduced which can be used to diagnose both the memory type of a time series, and the whitening filter type of a short memory time series. It is called prediction variance horizon function. To classify the model type of a time series, one uses the shape of PVH and the value of the horizon HOR (defined as the smallest value of h for which PVH(h) less than or = 0.05). The analysis of a real time series, called Freeze, is described.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jun 01, 1980
- Accession Number
- ADA094315
Entities
People
- Emanuel Parzen
Organizations
- Texas A&M University