Global Fit Package not holding coefficients constant
I am using the Global Fit Package to perform a global fit with a user-defined fit function and am experiencing a really puzzling issue: sometimes, when I check the "Hold?" box to hold a fit coefficient constant, the coefficient is still varied as if the box is unchecked. The problem is not consistently reproducible: sometimes unchecking and checking the "Hold?" box will fix the issue; other times it will not. This has occurred with both linked and unlinked coefficients. It does not happen with every coefficient (i.e. some of them are appropriately held constant while others are not).
I am running Igor Pro Version 9.0.0.10 (Build 37840) on Windows 11 Pro Version 23H2.
I am attaching a text document in which I pasted the global fit dialogue printed to the command window. I am globally fitting 41 traces to the same user-defined fit function, and there are 9 fit coefficients in the user-defined fit function. Some of the fit coefficients are linked between the different traces while others are not. In this example, the coefficients D and KB (unlinked between traces), and the coefficient OS (linked between all traces) were marked to be held constant. The first global fit attempt was successful, and D, KB, and OS were held constant. Then, I changed the initial guess for one of the D coefficients and hit the "Fit!" button again, still with all of the D, KB, and OS coefficients marked to be held constant. This time, the printed dialogue indicated problems with the fit:
*** Doing Global fit *** 9 iterations with no decrease in chi square
Hmm... Global Fit stopped for an unknown reason. Global fit results Fit stopped due to limit of iterations with no decrease in chi-square
and the OS coefficient and some (but not all) of the D and KB coefficients were varied, despite the "Hold?" box being checked.
Thank you in advance for any insight.
Hello Jack,
Thanks for writing in, but I'm sorry to hear about your trouble. The issue you describe sounds like a bug that is fixed in the latest Igor Pro 9.05 release, which you can get here: https://www.wavemetrics.com/software/igor-pro-9. You can also try out the very latest nightly build by going to Help->Igor Pro Nightly Builds from Igor. Please try either of these two options out and let us know how you get on.
Thanks,
Kris
October 9, 2024 at 10:01 am - Permalink
Hi Kris,
Thank you for your quick response. I am glad this bug has already been caught and fixed. Getting the update to Igor Pro 9.05 resolved the issue.
Thanks,
Jack
October 9, 2024 at 07:38 pm - Permalink
I'm so glad to hear it!
-Kris
October 10, 2024 at 08:37 am - Permalink