Fitting Points with Asymmetric Error


I googled for nearly 2 hours on this topic but got no positive answers. TGraph->Fit calculates chi2 using the low_error if curve is below the point and high_error if the curve is above the point. However, I’m not sure how the errors were assigned to MINUIT in the beginning and the low_error and high_error are used consistently from the beginning.

Besides all that, does anyone know of a robust and unbiased method of doing this? Thanks.


see documentation of TGraph::Fit at: