Hello,

Importance: This number is used only in a merge and only when the "Weight by importance" button in the Merge menu is checked. It's intention is to allow user-specified weights in a merge. If you trust one data set twice as much as another, the more trusted one might be given an importance of 2 while the other is given an importance of 1. This was in response to a feature request -- personally I am not clear how would know how to weight different data sets.

Oh, that's why it is called 'importance'. I actually used it for a different purpose: I have experimental data of an unknown sample but suppose it contains iron and probably iron carbide. I loaded experimental data for iron foil and iron carbide (I took the latter from the Newville-database - thanks for that ressource!) and merged them in chi(k). I used the 'importance'-factor for weighting the experimental values to reflect different concentrations. This should usually be done using Artemis - but I do not know if there exist crystallographic data for iron carbide, so that was the only way to simulate a mixture I could think of. Actually the comparison of the merged dataset with my unknown sample was not very good. Which might be explained by the fact I do not have a mixture of these substances. Or is there anything fundamentally wrong with what I was doing?

Edge step:

This does not appear to be working, if I check "Fix Step", enter a value into the edge step window, and then recompute the background, the value of the edge step reverts back to the automatic value.

I can affirm that bug: the edge step is not fixed just by choosing the 'fix step'-box. Work around: first click on one of the arrow buttons which will automatically enable 'fix step'. Then you can enter a value for the edge step which will now be fixed correctly. Cheers, Gerrit

On Sunday 05 June 2005 13:55, Gerrit Schmithals wrote:

Supplement: In the meantime I found crystallographic data for iron carbide, but I was also able to produce an excellent fit with the method I described by merging in Athena. So I believe merging of reference substances in Athena can provide a quick way to simulate a mixture.

Gerrit, That is an interesting use of the importance thing that had not occurred to me. It's certainly fun to see people doing things with my software that I hadn't thought of. You might take a look at the linear combination fitting dialog, which can be found in the Analysis menu. You can make arbitrary combinations of standards as well as fit them to the data. I have put a lot of work into that part of Athena recently and would be glad about all the bugs that people will no doubt find ;-) Also, you were right about the bug in the edge step and fix step controls. It was a little tricky to find because it was dependent on the order in which you clicked things. It should be working correctly in the next release. B -- Bruce Ravel ----------------------------------- bravel@anl.gov -or- ravel@phys.washington.edu Environmental Research Division, Building 203, Room E-165 Argonne National Laboratory phone and voice mail: (1) 630 252 5033 Argonne IL 60439, USA fax: (1) 630 252 9793 My homepage: http://feff.phys.washington.edu/~ravel EXAFS software: http://feff.phys.washington.edu/~ravel/software/exafs/

On Monday 06 June 2005 09:27, Gerrit Schmithals wrote:

How do I implement more than 4 references? There is a box for entering the number but it does not enlarge the reference-list as I expected it to do.

All of this is explained in the doc page for the LCF dialog. (That is, I think I explained it all there. It may be [is?] poorly explained ....;-) The length of the standards list is set in the preferences dialog. The relevant parameter is linearcombo->maxspectra. The huge new feature in the LCF dialog is what I am calling combinatoric fitting. Suppose you have 4 standards in the list in the LCF dialog. You can press the button that says "Fit all possible combinations" and Athena will cycle through all fits of 2, 3, or 4 of the standards to your data. Now suppose that you have 11 standards in the list. Cycling through all possible combinations would be extrememly time consuming. What's more, the fits involving many standards probably wouldn't be very believable in most situations. So you can tell Athena to limit the order of fits. By setting the widget below the list to 4, you tell Athena to fit all possible combination of 2, 3, or 4 out of the 11 possible standards. (FWIW, that's 550 fits -- it takes about 12 minutes on the computer I am sitting at right now.) There were also some problems in how the interpolating of the standards onto the energy grid for the fit was done in certain situations in previous versions of Athena. Interpolation should be working properly now. HTH, B -- Bruce Ravel ----------------------------------- bravel@anl.gov -or- ravel@phys.washington.edu Environmental Research Division, Building 203, Room E-165 Argonne National Laboratory phone and voice mail: (1) 630 252 5033 Argonne IL 60439, USA fax: (1) 630 252 9793 My homepage: http://feff.phys.washington.edu/~ravel EXAFS software: http://feff.phys.washington.edu/~ravel/software/exafs/

On Tuesday 07 June 2005 11:57, Gerrit Schmithals wrote:

I had an Athena project with some references loaded as chi(k)-data and some unknown samples as chi(k) as well. Athena won't let me start the Linear Fit-Dialog stating that it (she) needs at least three xmu-spectra. Changing the preference to "standard fit in  k-space" does not help. Workaround: I just loaded some dummy xmu-Spektra and can now work with the dialog.

Also, only these xmu-spectra can be selected from the "standards pluck boxes", the only way to insert the chi(k)-references is to use "use marked groups". That is a bit of a problem when you have to move one of the groups up because its value is negative. Workaround is to move up the groups in in Data group section.

Hmm... it seems that I had not done any testing with chi(k) records as standards. There are several problems. Thanks for pointing this out.

- Is there any possibility to plot the linear fit and the data in R-space right away? (without first saving it as a datagroup)

That can be done. Good idea.

- It would be nice to be able to store the relative weigths together with the newly created dataset.

I don't really understand this. The results of the LCF fit are saved as title lines for the new data group, but you must be looking for something else. B -- Bruce Ravel ----------------------------------- bravel@anl.gov -or- ravel@phys.washington.edu Environmental Research Division, Building 203, Room E-165 Argonne National Laboratory phone: (1) 630 252 5033 Argonne IL 60439, USA fax: (1) 630 252 9793 My homepage: http://feff.phys.washington.edu/~ravel EXAFS software: http://feff.phys.washington.edu/~ravel/software/exafs/

How about writing the relative intensities in % as well if not using "force to sum to 1"?

Doing so would presume a particular interpretation of the situation where the weights don't sum to one. You would expect the weights to sum to one if: 1. The standards and data are all calibrated and aligned consistently and that any e0 parameters used in the fit are well understood 2. The standards and data are all normalized consistently 3. The set of standards is representative of the contents of the unknown data If the best fit is one where the "sum to 1" constraint is lifted and the sum deviates considerably from 1, how do I know the reason a priori? So I (wearing my software developer hat) am hesitant to do what you suggest for fear of biasing the interpretation of someone else's data. BTW, thanks for yesterday's posting, Gerrit. It helped me find a pretty large number of bugs in the LCF dialog. B -- Bruce Ravel ----------------------------------- bravel@anl.gov -or- ravel@phys.washington.edu Environmental Research Division, Building 203, Room E-165 Argonne National Laboratory phone and voice mail: (1) 630 252 5033 Argonne IL 60439, USA fax: (1) 630 252 9793 My homepage: http://feff.phys.washington.edu/~ravel EXAFS software: http://feff.phys.washington.edu/~ravel/software/exafs/