R help - Mar 2009 - goodness of fit metrics for non-linear models

Hi,

I use nls to fit Gaussian curves to datasets that are expected to be
Gaussian-shaped:

    gauss.fit = nls(y ~ amp*exp(-0.5*(x-x0)^2/theVariance^2) + theNoise,
data = smooth, start = gauss.fit.start)

Some of these datasets are indeed shaped like Gaussians, while others
are not.  I would like to use a goodness of fit metric to assess whether
a Gaussian curve is a good fit to the data.  I wonder what metric would
be appropriate for this purpose.  I saw some discussions on this list
that suggested that R^2 is not meaningful in the non-linear regression
context, and that is why it's not reported in the nls object.  Are there
other, more appropriate goodness of fit measures?

Thanks.

Yury

	[[alternative HTML version deleted]]