When returning the TFitter::GetConfidenceIntervals in the special case of a TGraph, shouldn’t the scaling by the chi-square be omitted?
I followed the argument here for why this is included in general:
But for the TGraph documentation, it mentions that the error bars are already scaled by the chi-square (while the chi-square has not been rescaled similarly):
- When fitting a TGraph (i.e. no errors associated with each point), a correction is applied to the errors on the parameters with the following formula: errorp *= sqrt(chisquare/(ndf-1))