Consistency for the least squares estimator in nonlinear regression model

The consistency problems of the least-squares estimator [theta]n for parameter [theta] in nonlinear regression model are resolved perfectly. Assuming that the tth absolute moments of the model errors are finite, for t[greater-or-equal, slanted]2 and the errors satisfy general dependent conditions, we obtain the same probability inequality as that in Ivanov (Theory Probab. Appl. 21 (1976) 557) which has independent identically distributed errors; for 1<t<2, we first obtain weak consistency and weak consistency rate of [theta]n.