Hi there! I have a question about fitTo method of RooAbsPdf. I have to fit signal + background to data just like it is performed in the example here gitlab.cern.ch/will/TRooFit/-/blob/master/demo/abcdMultiBinExample.C
I want to know what fit to data exactly is, and how the method works in details? Maybe there is any additional literature on this topic? Is there any way to check if the result is right?
Also the estimates depend strongly on binning. e.g. the estimate of the number of background events is 300 when using 3 bins for the fit, 800 when using 20 bins, 900 when using 30 bins and it changes hardly when the number of bins gets greater. Is it correct and is there any way to find optimal binning?
I can attach the code if needed.
Here is an example of prefit and postfit histograms I get.
I would appreciate any advice.
Thank you in advance.
May be @moneta can help you.
is based on some code, TRooFit, which is an add-on to ROOT and I don’t know much about it.
fitTo in your case performs a maximum likelihood fit, probably a binned fit, to a combined data set of 4 different categories, representing the 4 different regions, A, B, C and D.
I guess you should use a proper binning that is smaller than the intrinsic variations of your distributions, 3 bins is certainly too small, but something larger than 30 is probably ok.
Thank you for your reply!
i.e. it is correct that there is such a strong dependence between the number of bins and estimated number of events and I just have to choose optimal binning to get the correct estimate?
You should not have this strong dependence. If you have it is because something is wrong, maybe because the binning does not make any sense.
Thank you very much for your help! It seems that I managed to solve this problem by using additional normalization coefficient for background.
This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.