Re: leasqr - covp

[Olaf Till]

On Fri, Dec 06, 2013 at 05:23:30AM -0800, ritik wrote:
> Hello,
> 
> I use leasqr to fit x-ray diffraction peaks. If I have more than 3
> parameters the covp matrix is NA. Is it possible to get the error of the fit
> somehow in the case of >3 parameters?
> Thanks in advance.

That is not necessarily a matter of the number of parameters. It is a
matter of how well conditioned certain matrices are, which again may
depend on how independent the parameters are from each other.

The function curvefit_stat (based on residmin_stat) should be better
in computing all computable elements of covp than leasqr, and still
better in computing all computable elements of corp.

Olaf

-- 
public key id EAFE0591, e.g. on x-hkp://pool.sks-keyservers.net

signature.asc
Description: Digital signature