[Ifeffit] Standard deviation of R value
anatoly.frenkel at stonybrook.edu
Fri Jan 11 10:18:03 CST 2019
As an additional comment, it is a murky issue whether the reported thing
together with R is the standard deviation. It is not easy to find it in the
literature but there are some papers dating back to the 1990s that state
that the uncertainties reported in the EXAFS fitting results correspond to
the 95% confidence level which means 2 standard deviations for an
approximately normal data set. I don't think it is correct. Regardless
whether the distribution of some or all fitting variables are treated by
the Levenburg-Marquartd method as normal or not, the reported error is
really not a standard deviation but a square root of the element in a
covariance matrix. What is the relationship of that and the standard
deviation of the fitting variable is one of the few remaining unsolved
mysteries in EXAFS analysis.
On Fri, Jan 11, 2019 at 11:05 AM Matt Newville <newville at cars.uchicago.edu>
> Hi Erin,
> On Fri, Jan 11, 2019 at 8:19 AM Erin Finley <efinley9033 at gmail.com> wrote:
>> Hello, I see in many publications the R value reported with a standard
>> deviation. However, when I look at my resulting data from a fit I don't see
>> one listed. I am still new to EXAFS research and so I am wondering where I
>> can get this information. I am assuming I need to calculate it from another
>> parameter I am just not sure which one.
> The uncertainties in R are typically in the fit report. But also, we
> typically refine a "delR" parameter that is R - Reff, where Reff is the
> distance (half path-lengh) for the scattering path. So, if you don't see
> an uncertainty reported for R but do see an uncertainty for delR (or
> whatever parameter name you used), use that. The uncertainty in R will
> (normally) be the uncertainty in delR.
> If I'm misunderstanding the question, pleas give more details about what
> you are doing.
> Thanks for your help!
>> Erin Finley
>> PhD Candidate
>> Brgoch Research Group
>> Department of Chemistry
>> University of Houston
>> Ifeffit mailing list
>> Ifeffit at millenia.cars.aps.anl.gov
>> Unsubscribe: http://millenia.cars.aps.anl.gov/mailman/options/ifeffit
> --Matt Newville <newville at cars.uchicago.edu> 630-252-0431
> Ifeffit mailing list
> Ifeffit at millenia.cars.aps.anl.gov
> Unsubscribe: http://millenia.cars.aps.anl.gov/mailman/options/ifeffit
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the Ifeffit