Mínimos cuadrados recursivos

Abstract

This paper presents the key features of the recursive least squares method emphasizing its usefulness for tests of structural break and autocorrelation. With regard to structural break, this method lends itself to a simpler class-room presentation than the more traditional Chow's F test. In connection with autocorrelation, the method's adventages and disadvantages vis-à-vis the Durbin Watson procedures are discussed. The paper also relates relates recursive least squares to the Kalman filter and provides some evidence to establish the superiority of this latter approach for generalized least squares estimation.