Simultaneous fit with TFractionFitter

Hi rooters,

I would like to use TFractionFitter to fit a data histogram as a sum of different Mc histograms. I would like to do a simultaneous fit of two (or more) different data histograms (different variables and binning) each one with its own set of Mc histogram components. I can do the same with a direct TMinuit application or with RooFit, but I can’t understand if it’s possible to do it with TFractionFitter. I saw some posts on the argument but I didn’t find any conclusion.

Thank you very much,
David Lopes

Hello David,

the short answer: no, I’m afraid this is not (presently) possible.

The longer answer: as you mention explicitly that the distributions fitted to
correspond to different variables, I conclude that the “posts on this
argument” you refer to most likely don’t apply to this case (they dealt
rather with fits to different data histograms involving the same template distributions).

So my guess is that the link between the different distributions is that the
fractions to be fitted are somehow related to each other – correct? A
priori, it does not seem very hard to extend the fit to do this. But I don’t
quite see how the relation (constraints) between the fractions in the
different fits could be formalized sufficiently to make such an extension a
general one (of course, if you can think of such a way this could be reconsidered…).


Hi Frank,

Thank you very much for your answer.

The fractions to be fitted are correlated, in the sense that they are the same for the two datasets (basically I try to fit uncorrelated variables for the same underlying process with the corresponding Mc histograms to put more “constraints” in the fit). With a chi2 fit (both in RooFit and directly in TMinuit) such a fit is straightforward.
Do you think that in this case an extension in TFractionFitter could be possible?

Thank you,