| T O P I C R E V I E W |
| sysyfun |
Posted - 05/07/2020 : 05:02:25 AM
I amusing OriginPro and trying to fit my curve, Nonlinear curve fitting (I builded new Funtion, y=a*b*x^(1/c)/(1+b*x^(1/c))). But when I successfully go through the process and hit fit until converge" then I keep getting this message: "fit did not converge, maximum iteration setting of 400 was reached". And then I tried again, but the same massage repeatedly. Please what do I do next to converge this fitting?
One more question, Is it okay to fit the parameters a, b, c arbitrarily?
If one of the parameter a, b, c values needs to be fixed, what should be considered?(standard error, etc.)
|
| 3 L A T E S T R E P L I E S (Newest First) |
| Serdar Ozbay |
Posted - 01/07/2024 : 4:59:55 PM
quote:
Originally posted by Chris D
A fitting doesn't always have to converge to achieve acceptable results. Visual inspection may lead one to conclude that the fitting is good enough regardless of whether or not it converged.
You can change the number of iterations or tolerance if you like in order to try to achieve convergence. You can do that via the "Advanced" control group in the "Settings" tab. You can also keep hitting the "Fit until converged" button to get the fitting to keep iterating. But neither will guarantee the fitting will ever converge.
More than anything else, the decision to fix one or more parameters depends on the context of the fitting. Does it make sense to fix a given parameter based on a priori knowledge of the subject?
Thanks,
Chris Drozdowski
Originlab Technical Support
I registered just to thank you. I was trying to get the fitting done and I finally did it by spamming the converge button thanks to you :) cheers. |
| cc261 |
Posted - 05/11/2020 : 12:32:53 AM
Hi, sysyfun, refer to the results below, based on the data picked out from your chart, the result is perfect, stable and unique:
Root of Mean Square Error (RMSE): 0.0219015694536506
Sum of Squared Residual: 0.0129513261023932
Correlation Coef. (R): 0.999884550116949
R-Square: 0.999769113562574
Parameter Best Estimate
---------- -------------
a -12.7757419807544
b -0.0910114202427086
c 2.67150664693673
 |
| Chris D |
Posted - 05/07/2020 : 08:25:04 AM
A fitting doesn't always have to converge to achieve acceptable results. Visual inspection may lead one to conclude that the fitting is good enough regardless of whether or not it converged.
You can change the number of iterations or tolerance if you like in order to try to achieve convergence. You can do that via the "Advanced" control group in the "Settings" tab. You can also keep hitting the "Fit until converged" button to get the fitting to keep iterating. But neither will guarantee the fitting will ever converge.
More than anything else, the decision to fix one or more parameters depends on the context of the fitting. Does it make sense to fix a given parameter based on a priori knowledge of the subject?
Thanks,
Chris Drozdowski
Originlab Technical Support
|
|
|
