[Ifeffit] Reporting the fitting data (uncertainty associated with N)

Matt Newville newville at cars.uchicago.edu
Tue Nov 4 06:24:23 CST 2014


The amplitude parameter is N*S02, so that S02 and N are completely
correlated.   If you

    set  N = 2
    guess S02 = 0.8

and the fit returns

   S02 = 0.75 +/- 0.05

Then the correct interpretation is that

  N*S02 = 2*(0.75 +/- 0.05) = 1.5 +/- 0.10

If you had asserted that S02 = 0.75 and vary N, then you should get

  N = (1.5 +/- 0.10) /  0.75 =  2 +/- 0.13333

The uncertainty in S02 and N cannot be separated for a single
spectrum.   The most common approach is to assert that S02 depends on
X-ray energy resolution (beamline setup) and absorbing atom, and use a
constant value for S02 derived from spectra from samples with a
well-known coordination environment (a "standard").

--Matt


On Tue, Nov 4, 2014 at 3:22 AM, pushkar shejwalkar
<pshejwalkar2004 at gmail.com> wrote:
> Dear All,
>           I have seen and also been told by few professors that when you
> report the fitting you must report the uncertainty associated with the
> measurement as evaluated by the program used for the fitting.
> Now, When I do the fitting I typically get the +/- values for E0, delR, ss
> and SO2. However, the N value we always put either 1,2 or such whole
> integers. Is there any way to find out the uncertainty associated with the
> coordination number (N used in XAFS equation)?
> If so how to find such uncertainty? can we see this in the log file
> generated after the fitting is over?
> Thank you very much for the help
>
> Best Regards,
> Pushkar Shejwalkar.
> Post-doctoral -Researcher,JSPS Fellow
> Hokkaido University,
> Sapporo,
> Japan
>
>
> _______________________________________________
> Ifeffit mailing list
> Ifeffit at millenia.cars.aps.anl.gov
> http://millenia.cars.aps.anl.gov/mailman/listinfo/ifeffit
>



More information about the Ifeffit mailing list