CI.tar.gz (10 KB)
Dear kind people,
I’m trying to calculating confidence interval. My data root file and macro are uploaded. I have no idea why my code gives unrealistic huge confidence interval? Anybody have any idea? I’m using root 6.10.08 Thank you very much.
I think there is a problem in using the covariance matrix from your fit due to numerical errors. The two parameters resulting from the fit are very correlated.
The first thing, I would suggest to rescale your data to have then in both x and w to be around 1 and not around 10^-3 in y and 10^3 in X
This will reduce the condition number of the covariance matrix
Yes, it works. Now I’ve had reasonable result. Thank you.
Is there any way to ultimate solution to avoid this kind of error? It’s a bit annoying to scale, fit, and rescale again.
I understand that this can be annoying, but unfortunately the computations are performed with a limited precision. In principle we could do this scaling automatically before fitting the data and rescale back the result, but we don’t have yet this option.
It is maybe something that we could add in the future
I see. If that option is included, it will be great.
Thank you very much for your kind anwser.
This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.