A Recursive Approach to Parameter Estimation in Regression and Time Series Models.
Abstract
This paper discusses the recursive (on line) estimation of parameters in regression and autoregressive integrated moving average (ARIMA) time series models. The approach which is adopted uses Kalman filtering techniques to calculate estimates recursively. This approach can be used for the case of constant as well as time varying parameters. In the first section the linear regression model is considered and recursive estimates of the parameters, both for constant and time varying parameters, are discussed. Since the stochastic model for the parameters over time will be rarely known, simplifying assumptions have to be made. In particular a random walk as a model for time varying parameters is assumed and it is shown how one can determine whether the parameters are constant or changing over time. In the second section the recursive estimation of parameters in ARIMA models is considered. If moving average terms are present, the model has to be linearized and the Extended Kalman Filter can be used to recursively update the parameter estimates. The first order moving average model is discussed in detail. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- May 01, 1978
- Accession Number
- ADA060717
Entities
People
- Johannes Ledolter
Organizations
- University of Wisconsin–Madison