THE ISERATION PROCEDURE AND ITS CONVERGENCE OF M ESTIMATOR IN NONLINEAR REGRESSION

Gauss-Newton iteration method and modified G-N method and their conyergence of M estimator in nonlinear regression are discussed in this paper. We generalized the resalts of Jennrich（6）and Gallant et. ale （4）from least squares estimator to M estimator. The numerical example shows that the modified G-N method is more preferable and the M estimator is very helpful to reduce the effects of outliers.