[R] se.fit in predict.nls

[Prof Brian Ripley]

I think you first need to understand how you would compute the standard 
errors.  Since the model is non-linear, it is hard to see how this could 
be done accurately.  If the parameters are estimated with negligible error 
compared to the non-linearity (and you can explore that since profiling 
gives you an idea of the variability), you can use local linearization to 
do this. If not, you can simulate.

On Thu, 19 Jan 2006, Manuel Gutierrez wrote:

> Sorry to be so persistent but I really need to get
> some measure of the error in the predictions of my nls
> model.

That is not what se.fit would tell you.  It gives you a measure of the 
variability in the fitted mean curve at one point.  If that is large 
compared to the variability about the true mean curve, you are likely to 
have trouble finding it (and the standard error may not an adequarte 
summary as it is likely to be an asymmetric distribution).

> I've tried to find out what predict.nls does and I've
> got down to
> MiModelo$m$predict
> function (newdata = list(), qr = FALSE)
> {
>    eval(form[[3]], as.list(newdata), env)
> }
> <environment: 0x88a076c>
> But I can not find what is "form".

You really do need to understand all the code, and lexical scoping.

> Any help, please.
> Manuel
>
>
> Manuel Gutierrez wrote:
>> The option se.fit in predict.nls is currently
> ignored.
>> Is there any other function available to calculate
> the
>> error in the predictions?
>> Thanks,
>> Manuel
>>
>
>
>
> ______________________________________________
> LLama Gratis a cualquier PC del Mundo.
> Llamadas a fijos y móviles desde 1 céntimo por minuto.
>
> ______________________________________________
> R-help at stat.math.ethz.ch mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html
>

-- 
Brian D. Ripley,                  ripley at stats.ox.ac.uk
Professor of Applied Statistics,  http://www.stats.ox.ac.uk/~ripley/
University of Oxford,             Tel:  +44 1865 272861 (self)
1 South Parks Road,                     +44 1865 272866 (PA)
Oxford OX1 3TG, UK                Fax:  +44 1865 272595