In linear regression the mean surface in sample space is a plane; in non-linear regression it may be an arbitrary curved surface but in all other respects the models are same. Fortunately in practice the mean surface in most non-linear regression models will be approximately planar in the region of highest likelihood, allowing some good approximations based on linear regression techniques to be used, but non-linear regression models can still present tricky computational and inferential problems.
KeywordsLinear Parameter Derivative Information Intrinsic Curvature Weight Loss Data Observe Information Matrix
Unable to display preview. Download preview PDF.
- 1.In S-PLUS 3.4 and earlier (and possibly later!) there is a bug which may be avoided if the order in which the parameters appear in the start vector is the same as the order in which they first appear in the model. It is as if the names attribute were ignored.Google Scholar
- 2.Included in S-PLUS 3.4 and later, and also available from the statlib archive as a library.Google Scholar
- 3.If nlme is being used as a library under Unix, these will be in library nlmedata .Google Scholar
- 4.Professor Douglas Bates has kindly permitted us to include his predict . nls and anovanis methods in our MASS library.Google Scholar
- 5.and differently from the help page which has information.Google Scholar
- 6.For this rather complicated-looking model specification, using deriv3 to produce a model function with gradient and hessian attributes may cause memory overflow problems on some machines.Google Scholar