Orthogonal Distance Regression vs Chi2

ROOT
1

Hi,

I’m fitting a simple line to a TGraphErrors with both x and y errors. The parameters and associated uncertainties result to be exactly the same as compared to a Orthogonal Distance Regression (carried out using the scipy odr module).

My question is: the fit chi2 value calculated by ROOT contemplates only y-errors or is it calculated also using x-errors ?

Thanks,
maffe

2

TGraph::Fit

3

By default ROOT uses for the TGrpahErrors::Fit both error on x and y, using a total least square as described in https://en.wikipedia.org/wiki/Total_least_squares

I think Orthogonal distance regression is slightly different, but maybe I am wrong and at the end the scipy implementation is the same as ROOT

Lorenzo