×

A note on the differential geometry of least squares estimation for nonlinear regression models. (English) Zbl 0830.62063

Summary: The sum of squares function of a nonlinear regression model is itself nonlinear in the unknown parameters and may therefore have more than one stationary point with respect to those parameters. This situation leads to complications of a computational and an inferential nature and is therefore of some interest.
Here, the relevant theory for characterizing observations on a nonlinear regression model for which the sum of squares functions has more than one stationary point is developed within a differential-geometric framework.

MSC:

62J02 General nonlinear regression
53A99 Classical differential geometry
PDFBibTeX XMLCite