| Author |
 Topic  |
|
|
zjsmcl
China
Posts |
 Posted - 01/13/2006 : 12:36:59 PM
|
Origin Version (Select Help-->About Origin):
when I use the NSLF to fitting the data, How to evaluate the Chi-sqr value.the less, the better? or is there any assess standard?just like the R-sqr,closer to 1,is the better? |
|
|
easwar
USA
1965 Posts |
Posted - 01/13/2006 : 3:36:36 PM
|
Hi,
What the iterative procedure in the fitter does is to minimize the chi-sqr value. The chi-sqr reported is the reduced chi-sqr, which is chi-sqr divided by number of degrees of freedom. You can review formula and further details in the help files.
The NLSF also computes and reports R^2. While fitting, you can simply go to script window and type
nlsf.cod=
to get the R^2 value.
At the end of the fit, you can go to the Action->Results menu in NLSF and click on PArameter Worksheet button to get a worksheet with all fit statistics.
Easwar
OriginLab
|
 |
|
|
zjsmcl
China
Posts |
Posted - 01/13/2006 : 6:26:43 PM
|
Thank you very much.
The question is,when I invoke different model to fit the data, after many iteration,the fitting curve simulate the data very good.at least looks very good.Can I draw a conclusion that this result is better than that one because its Chi-sqr is lower than that one and R sqr is higher than that one? or how to interpret the better?the two models' fitting cureve all look very good, only a small differences between the chi-sqr and r-sqr.
|
|
|
|
easwar
USA
1965 Posts |
Posted - 01/14/2006 : 11:37:22 AM
|
Hi,
If you want to compare two models to one dataset, there are statistical tests such as F-Test and AIC to do that. You could try this free tool from our File Exchange area:
http://www.originlab.com/FileExchange/details.aspx?fid=64
If on the other hand you want to compare fits to two datasets using one fitting function, you can use the Fit Comparison tool in Origin.
Easwar
Originlab
|
|
|
| |
Topic |
|
|
|
ask for help about NSLF