On Wed, 18 May 2011, Samantha Walker-Smith wrote:
> I've been using SPLAT to fit gaussians to my spectra and I was wondering
> how the errors involved in fitting a line centre and a FWHM are
> calculated, and whether the fitting takes into account the RMS of the
> line-free part of the spectra on its own? The errors in getting line
> widths just seem very small, especially when the spectrum is noisy and
> the line is only at the 3 sigma level. Thanks.
Hi Sam,
the errors are just derived from the data you identify as including the
line, so it is important that you include data out into the background.
The errors are derived from the covariances of the non-linear minisation
used. If your errors are too small, after including sufficient background,
that would suggest that the errors of your spectrum (assuming you have
some, that would require an NDF usually) are wrong, or in the absence of
errors that your data are correlated and the chisq value is too small.
Cheers,
Peter.
|