 |
 |
 |
 |
 |
 |
Systems and Control
CRES
Environmental Science
Research
Prospective Students
International Students
Amenities & Services
Accommodation
City & Area
Travel
Abstract
Young, P.C., (1999), Nonstationary time series analysis and forecasting, Progress in Environmental Science, 1, 1, 3-48.
| Until recently, the dominant paradigm in the analysis and forecasting of nonstationary time series has been the approach proposed originally by Box and Jenkins in 1970, which involves the en bloc processing of time series data that have been reduced to stationarity by pre-processing, using techniques such as differencing and nonlinear transformation. A more flexible and widely applicable alternative, which is now favoured in many different scientific disciplines, is to analyse the time series directly in their nonstationary form using recursive estimation and fixed interval smoothing. Here, the estimates of model parameters or state variables are updated sequentially, so allowing for the estimation of the time variable or state dependemt parameters that can be used to characterise models of nonstationary systems. This paper provides an introduction to the latest techniques in optimal recursive estimation and concentrates on the simplest class of models for nonstationary systems; namely time variable parameter, or 'dynamic', regression models, including linear (DLR), harmonic (DHR) and auto-regression (DAR), as well as the closely related time variable parameter version of the auto-regressive exogenous variables (DARX) model. In all cases, the utility of these methods is demonstrated through several practical examples. |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
|
|
Updating responsibility Arun Chotai. This page is copyright of Lancaster University.
12/10/01 - PGM.