[Ifeffit] Fitting Problems via Artemis

Matt Newville newville at cars.uchicago.edu
Mon Sep 27 15:12:46 CDT 2004

Hi Yu-Chuan,

I have not looked at the project file, but I have a few comments 

> I attached the apj file of magnesium orthovanadate. If you
> change the k-range, says, from 4~10 to 3~10. The fitting results
> will vary. Or you can change other stuffs. Somtimes the results
> are consistantly, somtimes not.

If you change the k range (or other FT parameters), the fitting
results can definitely change. How much the results change for
small changes in FT parameters is a good indication of the
reliability of the fit results.  Of course, saying that the
results change needs to include not only the best-fit values, but
also the uncertainties.
> By the way, I wonder if my first sell DW factor is about -0.0001
> and the uncertainties is +/-0.0015. Can I say that it is
> possible for positive DW factor for my first sell fitting? Thus,
> my fitting is reasonable?

Well, -0.0001 +/- 0.0015 is consistent with zero.  This is also a
very small absolute value for sigma2.  At k=10, the Debye-Waller
factor, exp(-2*k*k*sigma2) = 1.02 for sigma2=-0.0001.  There are
many ways to get a 2% error in amplitude.  One possibility 
(there are many) for why you could get a sigma2 that's smaller 
than "it should be" is that the fitting model puts in more static 
disorder than is really there.


More information about the Ifeffit mailing list