[R] standard errors in predict.nls ()

[Douglas Bates]

Jeremy.Whish at csiro.au writes:

> I am using r version 1.5.0 on a windows 2000 OS 
> 
> I have been converting an old S+ script I had that uses the nls function, to
> r. In S+ I could obtain the standard errors using predict(...,se = T)   In
> my documentation for R it says for predict.nls(), "At present `se.fit' and
> `interval' are ignored." my questions are 
> 
> 1. is this still the case  or am I just out of date 

Yes it is still the case.

> 2. is anybody working on these routines, if not could someone advise me on
> how to go about writing them.

I don't know of anyone working on them.  If you do not have a newdata
list (that is, you are predicting at the observed values of the
covariates) you should be able to get the standard errors of the
fitted values fairly easily from stored information.  I'll describe
that in private email.

If you do have newdata you will need to make a greater effort to get
standard errors.

Once you have the standard errors the rest is straightforward.



-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
r-help mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !)  To: r-help-request at stat.math.ethz.ch
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._