scientific article; zbMATH DE number 4115808

From MaRDI portal
Publication:4730669





The author analyses and generalizes two methods for estimating time varying parameter models: the regression method and the Kalmar filter- innovation correlation method. He shows in particular (i) that the identifiability conditions derived by applying the regression method coincide with the identifiability conditions derived by applying the Kalman filter method to the models considered, and (ii) that both regression and Kalman filter methods provide consistent and under certain circumstances asymptotically efficient estimates of the parameters of the model. Furthermore, an explicit estimator is derived for all the parameters of the model, including the initial values of the time-varying parameters and the covariance matrices of the model. Finally, by using the Kalman filter method, the relationship between the observability of the system and the identifiability of the parameters included in the state vector is examined, and it is shown that the identification condition and the observability condition coincide only if the smoothing equations are used to estimate the state vector.











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 Q4730669)