Chi2 fit with non diagonal variance matrix


I want to perform a chi2 fit to a histogram (histogram is a TH1 and the function a TF1). In the usual chi2 fit, it is assumed the histogram bin population is uncorrelated so, the covariance V matrix is diagonal.

However, I know that my in histogram the bins are correlated, so my V is not diagonal. Is there a way to instruct the TH1::Fit to use a non diagonal covariance matrix?


PS: the chi2 defined as: chi2 = (observed - expected)^T V^{-1} (observed - expected)

I think @moneta can help you with this question.

Hi Salva,

We don’t have in ROOT support for fitting an histogram or Graph with correlated data points. You will have to create yourself your Chi2 functions and minimise yourself.
The ROOT::Fit::Fitter class can help you doing this, as is shown in the tutorial line3Dfit.C, where in this case the distance line-point is minimised.


This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.