[R] What is the degrees of freedom in an nlme model

[Bert Gunter]

...

> > Aug;235(2):E97-102). The paper is freely available. The idea is to fit
> > the curves individually and obtain the residual sum of squares and
> > then fit the curves altogether somehow constraining some parameters
> > and then you have another residual sum of squares. Then you can do a
> > F-test. 

-- No you can't. This paper was apparently written by someone who doesn't
sufficiently understand the statistical issues. This is not uncommon -- even
papers in statistical journals sometimes get it wrong.


In my case, I wonder if I can use a mixed-effect modeling to
> > do the simultaneous fitting job.

-- You need to consult with your local statistician. This forum is not the
appropriate venue for difficult statistical questions that require intimate
familiarity with the data and an understanding of the scientific questions
of interest.

-- Bert Gunter
Genentech Nonclinical Statistics