Robust M-Procedures in Univariate Nonlinear Regression Models

摘要

In univariate nonlinear regression mo dels, estimator and teststatistics based on (generalized) least squares and maximum likelihood meth-ods are usually nonrobust; M-procedures are better in this respect. Ourproposed M-estimators, and M-tests are formulated along the lines of gener-alized least squares procedures and their (asymptotic) properties are studied.Computational algorithms are also considered along with