Book chapter
RECURSIVE ESTIMATION AND ADAPTIVE FORECASTING IN ARIMA MODELS WITH TIME VARYING COEFFICIENTS
Applied Time Series Analysis II, pp.449-471
Elsevier Inc
1981
DOI: 10.1016/B978-0-12-256420-8.50020-4
Abstract
Recursive estimation and adaptive forecasting in ARIMA models in which the coefficients themselves follow a stochastic process are discussed. In these models the influence of past observations on the coefficient estimates and on the forecasts is discounted.
The random walk as a model for time varying coefficients is studied in detail. In this model the covariance matrix of the innovations determines the variability of the coefficients. Maximum likelihood estimates for these, in practical applications unknown, parameters are derived. The estimates are then used to recursively update coefficients and forecasts. In a simulation study the forecast improvement for a first order autoregressive model with time varying coefficients is discussed.
Details
- Title: Subtitle
- RECURSIVE ESTIMATION AND ADAPTIVE FORECASTING IN ARIMA MODELS WITH TIME VARYING COEFFICIENTS
- Creators
- Johannes Ledolter - University of Iowa
- Resource Type
- Book chapter
- Publication Details
- Applied Time Series Analysis II, pp.449-471
- Publisher
- Elsevier Inc
- DOI
- 10.1016/B978-0-12-256420-8.50020-4
- Language
- English
- Date published
- 1981
- Academic Unit
- Statistics and Actuarial Science; Business Analytics
- Record Identifier
- 9984380597502771
Metrics
2 Record Views