[R-sig-ME] Port algorithm in nls

[Ken Beath]

On 19/06/2008, at 10:05 AM, Luis Orlindo Tedeschi wrote:

> Folks, I'd like to hear comments about a statement in the help and  
> manual of
> R 2.7 about the port algorithm in the nls function.
>
> It says: "The algorithm = "port" code appears unfinished, and does  
> not even
> check that the starting value is within the bounds. Use with caution,
> especially where bounds are supplied."
>
> IF, and only IF, it is unfinished how reliable are the estimates of  
> the
> parameters and why does R 2.7 have it?
>
> It sounds confusing to me... Mainly because most of my fittings are  
> done
> with port and I do use the bound statement. Sometimes port is faster  
> than
> Gauss-Newton... but now I am confused.
>
> I tried to get more info in Bates and Chambers (1993) but I could  
> not find
> it clearly.
>
> Any suggestions?
>

R is free software, so at times people will implement something but it  
may not be implemented as well as desired. It may still be useful to  
them and thus may be useful to others, as it is in your case, but is  
included with caveats. Check that the parameter estimates are at a  
maximum, always a good idea anyway.

Ken