Robust Nonlinear Regression

Anlr Database

This appendix presents the variables created as databases to store the different kinds of objects needed in nlr, and used in this book. There are two kind of databases supplied in nlr: sample data and lists of nonlinear functions of type nl.form objects. The latter include nonlinear regression models, robust loss functions, and nonlinear variance models.

The sample data are the real examples or simulated data used in the book. Mostly they follow nonlinear patterns, and will be used to fit appropriate nonlinear regression functions. Each data set may be fit with one or more nonlinear regression functions. Some of them include outliers, robust fits for which are discussed in the book.

Since nl.form is a general object that stores nonlinear functions, it is flexible enough to store different kind of algebraic functions. Several nonlinear function models are stored in nlr databases as lists of the nl.form object type. There are three kinds of nonlinear model stored in the nlr databases:

• nonlinear regression models, , that describe the data curve pattern
• loss functions , used in minimizing robust loss object functions
• heteroscedastic variance nonlinear functions.

Although these are all stored in the nl.form object, in detail nlr uses them in slightly different ways, as explained ...