| T O P I C R E V I E W |
| hauli |
Posted - 02/02/2009 : 08:05:34 AM
Origin Ver. and Service Release (Select Help-->About Origin): 7.5
Operating System: Windows XP
Dear all,
Origin allows to fit multiple datasets simultamnousely.
when I use the bulit in equation "LINE" for more then one dataset in one fitting run, I get the same standard errors for all my parameters A and the same standard errors for all parameters B.
This is a bug
Using another built in function, e.g. DoseResp I get different standard errors for all parameters.
Why is this?
very best regards
hauli |
| 1 L A T E S T R E P L I E S (Newest First) |
| minimax |
Posted - 02/03/2009 : 04:22:39 AM
Hi hauli,
It is not a bug.
It happens to be the same for all polynomial form functions, such as Line, when fitting XYY data (different Y, same X).
First of all, in Origin75 when fit multiple datasets simultamnousely, it performs a Global fit. (Global fit's benefit is that it can share parameters among datasets on fitting.)
When doing global fit, the fitter uses the formulas for evaluating errors that depend on the total number of parameters that vary and the total number of data points.
All polynomial form functions can be recognized as linear model.
For a linear model,
the covariance matrix (whose diagonal values are standard errors) of parameters is:
Now for an XYY data, the model is
the covariance matrix turns to be:
As you can see here, since X1 is equal to X2 (XYY data), you can easily get a same matrix for the top-left corner matrix and the bottom-right corner matrix.
That is to say, the errors of a parameter will be equal for each dataset.
You will not get a same value if you fit with an XYXY data, even it uses Line function.
Hope it helps.
PS: The newest Origin (v8.0 SR4) provides Independent Fit besides Global fit, which means to fit each dataset individually one by one.
For Independent Fit with XYY data, the error will be differnt then (but still in proportion).
You may take a try on the V8 demo if you like:
Origin8 Download
Max
OriginLab Technical Support |
|
|
