scientific article; zbMATH DE number 1069604

Authors

Publication date 5 February 1998

double integrated autoregressive models generalized random walks linearization nonlinear and non-stationary time series pseudo-linear models transfer function models

Mathematics Subject Classification ID

Time series, auto-correlation, regression, etc. in statistics (GARCH) (62M10) Inference from stochastic processes and prediction (62M20)

Review text

This chapter introduced an approach to nonlinear and non-stationary time series analysis for a fairly wide class of linear, time variable parameter (TVP) or nonlinear systems, an approach which is related to Maurice Priestley's own work in this area. The method theory presented here exploits recursive filtering and fixed interval smoothing algorithms to derive TVP linear model approximations to the nonlinear or non-stationary stochastic systems, on the basis of data obtained from the system during planned experiments or passive monitoring exercises. This TVP model includes the State Dependent type of Model (SDM) as a special case, and two particular SDM forms due to Priestley and the present author are discussed in detail. The methodology presented here has wide application potential. In its SDM form, it is useful as an off-line method for the detection and identification of nonlinearities in time series. In this latter context, it can be considered as a pre-processing procedure, in which the recursive filtering and smoothing algorithms are used for identifying the nonlinear model structure, prior to more efficient parameter estimation based on these identification results. This complete identification-estimation approach is best illustrated by a rainfall-flow example, where the identified soil moisture nonlinearity is incorporated in the model, whose parameters are then estimated using the statistically more efficient SRIV algorithm applied to the resulting constant parameter, bilinear transfer function model. However, the TVP estimation methodology is not limited to off-line analysis. Because of its fully recursive formulation, the off-line identification algorithms can be used to develop on-line procedures for the adaptive estimation, forecasting and control of nonlinear and non-stationary dynamic systems. Here, the recursive filtering algorithms used in the earlier off-line studies can form the basis for adaptive system design, so integrating the processes of systems analysis and synthesis in a rather useful and elegant manner. Finally, the identification and estimation procedures discussed in this chapter are tools that have been developed specifically by the author, over a number of years, for the data-based mechanistic modelling dynamic systems, particularly in relation to the scientific investigation of environmental systems. For the entire collection see [Zbl 0846.00010].

Recommendations

Cited in

(16)

This page was built for publication:

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4356566)