ON THE TREATMENT OF CERTAIN RECURRENT NON-LINEARITIES IN REGRESSION ANALYSIS*

The model in (1) was first investigated by Cochrane and Orcutt [1], who, however, did not produce an analytic solution. A two stage procedure is suggested by Durbin [2] in the context of a far more general investigation. The estimating procedure as exposited in [2] is rather complex but it is shown to be consistent and asymptotically efficient. In the face of the specification (la) it is not, however, a maximum likelihood estimator.