Dear root users!

I have a small question related to the GraphFitChisquare function which is the default fitting function to be used by the TGraph::Fit() function. I’m doing a simple fit of a function (ax + b(1/x)) to 5 experimental values with both x and y errors. When using the TGraph::Fit() function I get values for a and b that differ from my own calculations when using a Maximum Likelihood technique and the method of effective variance. How is this done by the GraphFitChisquare function if it is not done by the method of effective variance?

Best regards

Bjarte Mohn

The method with “Effective variance” is implemented in the development
version in CVS (Thanks Anna Kreshuk).
See comments at: