| T O P I C R E V I E W |
| erickzhou |
Posted - 07/15/2010 : 09:08:00 AM
Origin Ver 8. and Service Release (Select Help-->About Origin):
Operating System:winXP
I use "nonlinear surface fit" to extract the 2-D polynominal trend of my data(256*256 data points). I want to know if the error of trend fitting is worse in boundary than other position. If it is true, how can I deel with it.
Your early reply will be appreciate, thank you1 |
| 4 L A T E S T R E P L I E S (Newest First) |
| larry_lan |
Posted - 07/16/2010 : 7:21:45 PM
The more data you have, the more "power" you have to make statistical inference. There are many values you can look into to evaluate the goodness of your fit.
Thanks
Larry |
| erickzhou |
Posted - 07/16/2010 : 10:28:17 AM
Thank you! I have another question. I was told that if the data points for surface fitting is enough, the boundary error would be small. However, if the data points for surface fitting is not enough, what kind of error will emergy beside the boundary error, i.e. what factor decided the accurancy of fitting results. |
| larry_lan |
Posted - 07/15/2010 : 11:17:04 PM
Hi:
We don't support weighting in Nonlinear Surface Fitting, but you can use Nonlinear Curve Fitting instead. Surface Fitting is a special case of Nonlinear Curve Fitting (Two Independent Variables). You just need to defined a two independent variables fitting function in any Category except "Surface Fitting". The Fitting Function Organizer can help to create new fitting functions. And you can also Share a 2D fitting function in "Surface Fitting" category to any other categories.
If data on the boundary have large deviation from the fitted surface, maybe you can use Statistical weight.
Choosing a proper model and diagnose fitted result is not an easy job, maybe the following references can help:
- Bruce Bowerman, Richard T. O'Connell. 1997. "Applied Statistics: Improving Business Processes". The McGraw-Hill Companies, Inc.
- Sanford Weisberg. 2005. "Applied Linear Regression". Third Edition. John Wiley & Son, Inc., Hoboken, New Jersey.
- William H. Press et al. 2002. "Numerical Recipes in C++", 2nd ed. Cambridge University Press: New York.
- Marko Ledvij. "Curve Fitting Made Easy". The Industrial Physicist. Apr./May 2003. 9:24-27.
Thanks
Larry |
| erickzhou |
Posted - 07/15/2010 : 9:36:07 PM
Some one told me that I can increase the weight of the boundary data points to overcome the problem of boundary error. But I don't know how to set the weight of data points in Origin 8, and I also don't know what weight should be set for each data point.
They also told me that if I have hundreds plus hundreds data points, the boundary error can be omitted. Is it true?
|
|
|
