| T O P I C R E V I E W |
| WildEnte |
Posted - 03/09/2004 : 11:32:48 AM
Hola!
I searched the forums & the Origin help, but found nothing, so here I go:
I fit a gaussian to a dataset (countrates) describing a gaussian - no problem so far. Now I choose "Statistical weighting" - also no problem. I get the same result if I choose an arbitrary dataset for the weighting where I put in the sqrt(data).
But what does Origin do when the datapoint is 0? The weighting is w=1/sqrt(0) in that case, but this must be handled in a different way because otherwise I'd get a a division by zero.
I need to know this... the chi^2 depends on the value Origin puts in there (chi^2 being basically the sum over the distances between fit and datapoints normalized to the error of the according datapoints).
Thanks for your input! |
| 2 L A T E S T R E P L I E S (Newest First) |
| WildEnte |
Posted - 03/10/2004 : 03:52:03 AM
Thanks for your quick and helpful reply. |
| easwar |
Posted - 03/09/2004 : 10:59:24 PM
Hello,
In versions prior to 7SR4, when a weighting method encountered a possible division by zero, Origin replaced the error by a smaller non-zero error from a neighboring point. We decided that was not appropriate to do and since 7SR4 we simply bring up a dialog stating that zero or extremely small values of error which are not consistent with the weighting method were found, and the fit does not proceed.
I presume you are using an Origin version prior to ver 7SR4 since you do not see the error message. I would suggest eliminating those points from your fit using the masking tool, or yourself replacing the zero error (if error comes from a different column than the y data) by what you may decide to be a suitable non-zero error value.
Easwar
OriginLab.
Edited by - easwar on 03/09/2004 11:02:35 PM |
|
|
