Non-stationary parameter estimation for small sample situations: A comparison of methods
- 30 March 1976
- journal article
- research article
- Published by Taylor & Francis in International Journal of Systems Science
- Vol. 7 (3) , 257-275
- https://doi.org/10.1080/00207727608941915
Abstract
This paper reviews eight methods for estimating the form of non-stationary parameter variation from input-output data when the parameter variation model is unknown. The methods discussed are based upon error bounding and adaptive estimation, and assessment of the relative efficacy of each method is carried out by the use of simulation experiments with two models of parameter variation: deterministic parameter changes, and an autoregressive model of parameter changes. It is concluded from this analysis that estimation based upon exponential data weighting, the random walk model, and Sage-Husa filter will usually be best in a variety of situations.Keywords
This publication has 17 references indexed in Scilit:
- Process identification for time series modelling in urban and regional planningRegional Studies, 1974
- Kalman filtering of systems with parameter uncertainties—a surveyInternational Journal of Control, 1973
- A Bayesian Approach to Short-term ForecastingJournal of the Operational Research Society, 1971
- System identification—A surveyAutomatica, 1971
- Adaptive filteringAutomatica, 1969
- An innovations approach to least-squares estimation--Part I: Linear filtering in additive white noiseIEEE Transactions on Automatic Control, 1968
- Optimal Non-stationary Estimation of the Parameters of a Linear System with Gaussian Inputs †Journal of Electronics and Control, 1963
- A New Approach to Linear Filtering and Prediction ProblemsJournal of Basic Engineering, 1960
- The Estimation of Distributed LagsEconometrica, 1958
- On the Statistical Treatment of Linear Stochastic Difference EquationsEconometrica, 1943