| Author |
 Topic  |
|
|
joelleoosterman
Canada
2 Posts |
 Posted - 02/22/2014 : 7:53:52 PM
|
Origin Ver. and Service Release (Select Help-->About Origin):
Operating System:
Dear OriginPro,
I have been trying to fit my data using non-linear curve fitting with a custom-made cosine function: y=M+A*cos(2*pi (x-H)/P)) in which M=measor, A=amplitude, H=phase, and P=period.
I have a few questions about the fitting results, and I hope someone can help me out with this. (I have read many help-topics and forum posts, but I'm still confused)
- In the parameter table, all values are given for the unknowns, including standard error. Is this standard error of the mean (SEM) or standard deviation? Can I use these errors for a direct comparison between two fitted curves? (e.g. compare the period of one fitted curve to the period of another fitted curve using a t-test?)
- When generating the curve, I have to enter parameter values. These can be fixed (boxed checked) or not fixed. Is it better to enter a fixed value? I feel that I'm biasing the data by doing so, but not sure.
- In the ANOVA table, what does the prob>F value mean? Does a value of >0.05 mean that there is no significant difference from the fitted curve?
Thank you so much for your help!
Joelle
Joelle |
|
|
lkb0221
China
497 Posts |
Posted - 02/24/2014 : 10:04:40 AM
|
Hi,
1). Standard Errors means SEM. You can see more details about our fitting algorithms in the following page:
http://www.originlab.com/doc/Origin-Help/NLFit-Algorithm
And we recommend you to use "Analysis: Fitting: Compare Datasets" to do comparison between two fitted curves by the same function.
2). Initial value the parameters will help you to get a better fit easier and faster, and prevent over-parameter.
3). "prob>F" is p-value. The smaller it is, the larger the significance.
Zheng
OriginLab |
 |
|
|
joelleoosterman
Canada
2 Posts |
Posted - 02/24/2014 : 10:59:41 AM
|
Hi Zheng,
thank you so much for your reply, that was really helpful.
Just a follow-up question: I have tried to compare two fits with each other using analysis\compare datasets. The outcome is "There is not enough information to draw conclusion". Both curves fit with a R2>0.90. I have read that the negative or missing F-value may cause this problem, but I don't understand how I can change this. Can you help me with this?
Then, for the p-value. Both of my fitted curves show an R2 of >0.90, but the p-value is 1. How is this possible?
Thanks again for your help, much appreciated.
Joelle
Joelle |
|
|
| |
Topic |
|
|
|
non-linear curve fitting