Two-stage IV-based estimation and parametrization selection for linear multivariable identification

Abstract

Previous work of the authors on the use of overlapping parametrizations of finitedifference equations in the identification of multivariable systems is continued. By considering numerical and accuracy issues, a two-stage, IV-based, joint parametrization-selection and parameter-estimation scheme is proposed. The parametrization selection seeks to ensure that the estimation problems are numerically well conditioned, whereas the second stage ensures statistical efficiency in certain circumstances.