[Ifeffit] error

Matt Newville newville at cars.uchicago.edu
Tue Jan 11 20:36:53 CST 2011


Hi JeongEunSuk,

2011/1/11 JeongEunSuk <eunsuk1986 at hotmail.com>:
> Hi all
> I fitted ZnMgO to know whether Mg is sited on Zn position.
> the method of fitting as following;
>   1st : to fit with some Mg located on Zn site.
>   2nd: to fit with some Mg which position is different from Zn cite
> The parameters for Mg like bonding length, disorder and others were set.
> When I was fitting, there are only 8 parameters in both two method. So that
> independent data point and variables are same.
>
> Both 1st and 2nd had same R-factor but reduced chi-square of 1st was lager
> than that of 2nd.
> It has been known that R-factor is independent of error( uncertainty in the
> measurement) but reduced
> chi-squrare is inverse proportional to error^2.
> So I think that same R-factor and different reduced chi-square for two
> method mean the fittings of two method are not different,
> but error used in reduced chi-square is different.
> I want to know how the error used in reduced chi-square is decided in FEFFIT
> code.

The details of the fit statistics have been explained in many places.

Since your questions are about numerical values being "the same" and
"different", it would be helpful to see some of the numerical values
from a log file or project file.

--Matt



More information about the Ifeffit mailing list