Constraints when using Global Fitting

Share icon

Thu, 05/26/2011 - 12:19 pm

I used the Global fitting package and found the following issues about constraints. I use Igor 6.22 on Mac OS X (10.5.8) Typically I have 10+ fitting parameters.

1. It causes problem when I set constraints such as: K1 < K2 < K3. K1, K2 and K3 are all fitting parameters.

2. The fitting program is easy to crash when too many constraints are set, which gives the following message, " Global Fit stopped for an unknown reason.". It runs well if I remove the constraints.

Any ideas on how to solve these issues?

thanks,

Yaohua
yaohualiu wrote:
1. It causes problem when I set constraints such as: K1 < K2 < K3. K1, K2 and K3 are all fitting parameters.

That is not a legal constraint expression. You need to break it into two expressions:
K1 < K2
K2 < K3

You need to read more about how constraints work in regular curve fitting. You can read the help files by executing this command on Igor's command line:
DisplayHelpTopic "Fitting with Constraints"
Quote:
2. The fitting program is easy to crash when too many constraints are set, which gives the following message, " Global Fit stopped for an unknown reason.". It runs well if I remove the constraints.

By "crash" I presume you mean that the fit ended with an error instead of a successful fit. It is easy to create a situation that is hard to solve when you use constraints. It is not essential, but is usually a good idea to make sure your initial guesses comply with the constraints.
If you are not using the latest update of Igor (6.22a) I highly recommend updating. Recently I added better error reporting for Global Fit.

John Weeks
WaveMetrics, Inc.
support@wavemetrics.com

May 27, 2011 at 08:46 am - Permalink