[R] nls.profile

[Douglas Bates]

On Thu, Mar 31, 2011 at 4:02 AM, Daniel Kaschek
<daniel.kaschek at physik.uni-freiburg.de> wrote:
> Hello,

> I use nls.profile to compute confidence intervals of parameter estimates
> of a non-linear model. When computing the profiles, the model function
> produces an error for certain parameter combinations. Therefore nls
> fails and so does nls.profile.

> Is there a way to tell nls.profile to ignore errors and to proceed with
> the next parameter value?

Not easily.  What I tend to do is to adjust (increase) the alphamax
parameter (which is a misnomer, it should have been alphamin but I got
confused) until the profile can succeed.