| Author |
 Topic  |
|
|
rainbowkoala
29 Posts |
 Posted - 09/23/2013 : 10:56:29 AM
|
Origin Ver. and Service Release (Select Help-->About Origin): OriginPro 8.5.0 SR1
Operating System: win7
My data is in the insert image.
When using "Apparent Fit", I can get a better fitting result, however, we get the fitting result about logy and logx.
When using "NLFit", I would get a worse fitting result,because NLFit usually gives numerically minimum chi-square, which means larger (X or Y) coordinate has more significant contribution to the fitting result.
Then how can I get a better NLFit fitting result about y and x£¿
Please find more details in the insert image.
Thanks for your help.
 |
|
|
Hideo Fujii
USA
1582 Posts |
Posted - 09/23/2013 : 5:39:46 PM
|
Hi rainbowkoala,
As far as sticking to the polynomial(when the power>1 in y=Sigma(x^power)), I think the result like in your lower
graph is at its "best" in terms of the Chi^2.
When the data is plot in linear scale like Graph1 below, the power of X in the function form seems
less than 1 like the square root. Indeed, if you free the power as a variable, say like Allometric1 (Y=a*X^b),
you get 0.3675 as in the screenshot (Graph1 and Graph2). By choosing other power function like BINeld,
you can get even better fitting as in Graph3 and Graph4. Of course, it's up to you to decide which type
of function you want to choose.
I suggest you reconsider the model of your data.
Hope this helps.
--Hideo Fujii
OriginLab |
 |
|
| |
Topic |
|
|
|
How to get a better NLFit fitting result£¿