| Author |
 Topic  |
|
|
Younger2008
USA
Posts |
 Posted - 02/17/2005 : 08:13:09 AM
|
Origin Version (7.5):
Operating System: MSOS 2000 SP4
I have several datasets with x1,y1;x2,y2; etc. When I use NonLinear Curve Fitting in Origin, for single curve fitting, there is a result log showing R^2, and the meaning R^2 is clear.
But when I select 'Fit Multiple Datasets' to fit three curves in the mean time, the results log only gives one R^2, not three corresponding to each curve. What is the meaning of this R^2?
And normally, the R^2 from 'Fit Multiple Datasets' is much better than that of single curve fitting. What is the difference of these two R^2?
Best regards,
Younger |
|
|
easwar
USA
1965 Posts |
Posted - 02/17/2005 : 08:58:05 AM
|
Hi Younger,
When you fit with multiple datasets (in the NLSF tool), all datasets are fitted simultaneously with the chosen function, in one iterative procedure. This is why you get only one R^2, one Chi-sq etc reported for the fit. In other words, the minimization is performed in one parameter space that is a combination of all the parameters for all datasets.
This procedure is best suited when you really want to fit multiple datasets by sharing some parameter, such as say fitting multiple gaussians where the centroid of the peak in each dataset is shared because for some physical reason it is known before hand that the peaks should all be at the same location, for example.
Currently with the multiple dataset choice, there is no option to instead tell NLSF to fit each dataset separately. We will add this option in a future version.
I also want to point out that the above is true for the NLSF tool only. For tools such as LR, PR there is a check box called "Fit all curves". If this check box is checked, then each dataset plotted in the graph is fitted individually.
So for now, with the NLSF, if you do not intend or do not need to share parameters, it is best to fit each dataset separately, that is one dataset at a time. One can write scripts/Origin C code to make this procedure easier if you have many datasets. See the file exchange tool by Mike Buess for example:
http://www.originlab.com/fileexchange/details.aspx?fid=56
Easwar
OriginLab
|
 |
|
|
Younger2008
USA
Posts |
Posted - 02/17/2005 : 10:18:18 AM
|
| Thanks four kind reply! |
|
|
| |
Topic |
|
|
|
Questions about NonLinear Curve Fitting