| Author |
 Topic  |
|
|
eemian
Germany
1 Posts |
 Posted - 10/25/2002 : 05:50:54 AM
|
Hi,
for the correct determination of uncertainties in our measurements Im interested in the procedure which is applied in ORIGIN for the determination of the uncertainties of the fitted parameters during curve fitting (linear and non-linear).
Does the procedure consider the uncertainties both of the x- and the y-values? Or does the program only use the pair of x,y-values without their uncertainties?
Can someone give me a reference where I can get information about the used numeric procedures?
Thank you in advance
Detlev |
|
|
easwar
USA
1964 Posts |
Posted - 10/25/2002 : 12:05:03 PM
|
Hi Detlev,
For both linear and nonlinear curve fitting, only error bars (uncertainities) in y are considered. Error bars in x are not considered. As for y error bars, you need to specify that they are to be used in the fitting process. In the Linear Fit tool, there is an "Error as Weight" check box which when checked will use the y error bars in the regression calculation. The error weighting is assumed in this case to be 1/yerr^2. In the case of the NLSF nonlinear fitting tool, the error weighting can be one of many choices such as statistical, instrumental etc.
Now, the error (uncertainities) in the fit parameters for both linear and nonliner regression that are reported are the standard errors from the computation. As for linear regression, the formula for the standard error can be found in Numerical Recipes in C, Second Edition, Page 663. As for nonlinear regression, the expression for standard error is documented in our help file under the Curve Fitting chapter. You can search for covariance in the help file and then look under the "Fitting Session Concepts" topic.
Easwar
OriginLab.
Edited by - easwar on 10/25/2002 2:58:09 PM |
 |
|
|
Mike
USA
357 Posts |
Posted - 10/25/2002 : 5:23:58 PM
|
Hi Detlev,
Easwar pointed out that we had neglected to include some information that might be helpful to you from our most recent product documentation. I have posted an updated .PDF file on curve fitting on our website. Specifically, you may want to look at the section entitled "A Note on Linear Fitting Theory."
I hope that this helps.
Mike
OriginLab |
|
|
| |
Topic |
|
|
|
uncertainties and curve fitting