Estimación recursiva de modelos lineales con restricciones entre los parámetros

Abstract

In this paper we show that reeursive estimates of a linearly constrained model parameters can be obtained by initializing adequately the standard method (reeursive least squares). The advantage of this approach with regard to it alternative (reduced-dimension filters) are 1) the same algorithm can be used for estimating constrained and unconstrained models and 2) it allows a recursive constraint testing. The theoretical analysis is completed with an example, which shows how the reeursive estimation provides useful insights about the parameter stability and the effeet of each observation over the estimates.