[Ifeffit] Multiple k-weight fitting

Bruce Ravel bravel at anl.gov
Mon Jan 22 12:12:06 CST 2007

On Monday 22 January 2007 11:50, David C. Schmitter wrote:
> Hi everyone.  I am currently trying to reduce the correlations between N
> and sigma squared.  I have been told that multiple k-weight fitting is the
> way to go about this; but I am curious what the best method is to try and
> do this using Artemis.   Is it simply a matter of clicking the other
> weighting options available?  If so, what should I be looking for to
> determine whether my parameters are correct or not?
> I am not sure if this is a redundant question in terms of the mailing list;
> but I am a relatively new user and cannot remember seeing anything about it
> yet.

Hi Dave,

The intention of the mailing list is to have a forum for asking these
sorts of questions.

Multiple k-weighting is one of the standard methods in Ifeffit to
attempt to introduce some information that uncorrelates N and sigma^2
(or deltaE and deltaR, for that matter).

I recently gave a talk that touched upon this topic.  Scroll to the
bottom of this page: http://cars9.uchicago.edu/~ravel/talks/ and
forgive the typos.  (It was written on an airplane.)

In short, multiple k-weights is very easy to introduce in artemis.
Simply click more than one of the items in the the k-weight box in
Artemis' data view (which is the one that shows up when the program is
first launched).  In that talk, I explain both how multiple
k-weighting works in Ifeffit's fitting algorithm and give some hints
as to why it works.

You might also take a look at the tutorials by Matt and Shelly in the
tutorials section of xafs.org.  I think they both touch on this topic
as well.

As a final note, I should give the caveat that multiple k-weights and
other tricks may help to reduce the correlations, but -- except in
special situations -- propbably won't make them disappear.  The
amplitude parameters will always be correlated at some level, as will
the phase parameters.  It's just the nature of the problem.  Even in a
good, believable fit to a very simple problem like a metal foil, the
correlations might be in the 80s.


 Bruce Ravel  ---------------------------------------------- bravel at anl.gov

 Molecular Environmental Science Group, Building 203, Room E-165
 MRCAT, Sector 10, Advanced Photon Source, Building 433, Room B007

 Argonne National Laboratory         phone and voice mail: (1) 630 252 5033
 Argonne IL 60439, USA                                fax: (1) 630 252 9793

 My homepage:    http://cars9.uchicago.edu/~ravel 
 EXAFS software: http://cars9.uchicago.edu/~ravel/software/exafs/

More information about the Ifeffit mailing list