| T O P I C R E V I E W |
| jd0331 |
Posted - 10/02/2009 : 6:59:06 PM
Origin Ver. and Service Release (Select Help-->About Origin): Origin 8 SR1
Operating System: Windows XP
Hi, I tried NLFit with using Origin 8, and would like to know what the 'Tolerance' is.
I could not find any information regarding the Tolerance value on the web.
You can see it if you follw these steps.
NLFit->Setting->Advanced->Fit Control->Iterations->Tolerance
I figured out that I can change the tolerance value before starting NLFit, and the results were better when I set the value with a smaller number.
What I want to know is what the Number is.
Is it a fractional or percentage allowable error?
If not, what is that?
If you could let it to me with a equation, it will be greatly appreciated. |
| 3 L A T E S T R E P L I E S (Newest First) |
| easwar |
Posted - 10/05/2009 : 3:12:17 PM
Hi,
I apologize, it is the absolute value of difference divided by the sum....and it is the reduced_chi-sqr but then reduced_chi-sqr is just chisqr/degree_of_freedom, so that is a common factor which would cancel out anyway.
Easwar
OriginLab
|
| jd0331 |
Posted - 10/05/2009 : 2:17:45 PM
Thanks!
It seems that the tolerance value is negative if the equation that you formulated is correct because the chi^2 usually decreases with increasing the number of the iteration.
Isn't it?
By the way, is the chi^2 that you formulated a reduced chi^2 or chi^2 ?
quote:
Originally posted by easwar
Hi,
The equation is:
Tolerance = (new_chisqr - old_chisqr)/(new_chisqr + old_chisqr)
You are correct, this was not documented, we just added to our documentation for the next version.
The iterative procedure looks at both Max Iterations and Tolerance. Any time the tolerance is met within the specified max iterations, the fitting stops. And if tolerance is not met after specified max iterations, then the fitting stops as well. So you may need to adjust both numbers depending on your data and model. If you are having a problem with a specific model and dataset, you can send them into tech support.
Easwar
OriginLab
|
| easwar |
Posted - 10/05/2009 : 09:52:09 AM
Hi,
The equation is:
Tolerance = (new_chisqr - old_chisqr)/(new_chisqr + old_chisqr)
You are correct, this was not documented, we just added to our documentation for the next version.
The iterative procedure looks at both Max Iterations and Tolerance. Any time the tolerance is met within the specified max iterations, the fitting stops. And if tolerance is not met after specified max iterations, then the fitting stops as well. So you may need to adjust both numbers depending on your data and model. If you are having a problem with a specific model and dataset, you can send them into tech support.
Easwar
OriginLab |
|
|
