On Thu, Jan 30, 2025 at 3:49 AM Otal Eugenio via Ifeffit < ifeffit@millenia.cars.aps.anl.gov> wrote:
Hi all, I am experiencing an issue with CN and sigma2, which are correlated. My CN value is too high, and I would like to reduce it to see if I can improve the fit. I have three questions: Are constraints a viable option? If not, could you clarify ZjQcmQRYFpfptBannerStart This Message Is From an External Sender This message came from outside your organization.
ZjQcmQRYFpfptBannerEnd
Hi all,
I am experiencing an issue with CN and sigma2, which are correlated. My CN value is too high, and I would like to reduce it to see if I can improve the fit.
For sure, N and sigma2 are correlated. This does not mean their values are unreliable. Each value will also have a reported uncertainty. That uncertainty accounts for the correlation. There are several factors that directly affect EXAFS amplitudes, including normalization and having the right species of scattering atom. Since we don't know anything about what you are doing, it would be hard for us to be sure about such things. I have three questions:
1. Are constraints a viable option? If not, could you clarify the alternatives?
What is the quantity you would like to constrain? I can't immediately think of a physically meaningful connection between N and sigma2. There can be such connections between N and R.
1. The examples I found are in Python, but I am using the GUI. How should I correctly define a constraint?
Path Parameters can be written as mathematical expressions of variables. You can use variables in expressions for multiple Path Parameters. That's a constraint.
1. GPT suggested the following: max(4, 12 * exp(-sigma2_O232 / 0.02)), is it correct?
Hm I don't see where that comes from. Why should N depend on sigma2?
1. I attempted to apply this constraint, but now I cannot delete it. Restarting Larch for each value is not a practical solution. How can I modidy the constraint efficiently?
In Larix, each Path for Feffit gets its own "Notebook Page/Tab" with entries to enter the mathematical formula for each Path Parameter. Each named variable in those expressions (that is not a recognized function name) gets put onto the Parameters tab where it can get an initial value, bounds, whether it is to be Varied, Fixed, or Constrained, and so forth. If constrained you enter a mathematical formula for that value in terms of the other variables. That named Parameter will then change value during the fit, but it will not vary independently. It will follow the constraint expression you provide. If a Parameter in the Parameter list is not used by any Path Parameter, it will be set to be Skipped when the Feffit model is calculated or a fit is run. You can click on "edit Parameters" -- it will list all the Parameters, selecting the unused ones, and you can delete them. It should not be necessary to delete them, but you can do that. --Matt