Consistent estimation for some nonlinear errors-in-variables models

The issues of identification and estimation of nonlinear errors-in-variables models are explored. The deficiencies of the conventional definition of identification are discussed and an alternative definition in terms of observed samples is suggested. The consistency and asymptotic normality of minimum distance estimators are studied. To simplify the computation, a two-step estimation procedure is also suggested in which a consistent estimate of a subset of parameters is first obtained and treated as if they were known in the second-step estimation of the rest of the parameters. Conditions for the consistency of this two-step estimator and its asymptotic variance-covariance matrix are also derived.