Nonlinear regression models: computation of estimators and curvatures

Abstract

This book is oriented toward statistical properties of estimators, so that methods of numerical computation of estimates are not our focus of interest. Moreover, such methods are contained in many statistical packages. Nevertheless, a short chapter on the topic is necessary to complete the exposition, and to see the connection with the geometric interpretation of the regression model. This connection is especially clear in the Gauss-Newton method. This is a classical method, however, as stated in [Rt], the method is still advisable in applications. The reader interested in more detailed information on the various computational methods is referred to [KG].