I have a set of data that I would like to fit to a model constructed from two histograms, one representing background, and another representing the signal. The histogram for the background is constructed from a data sideband region and the signal histogram is constructed from a Monte Carlo sample. I am curious about how RooFit will handle two aspects of this fit? The first is the statistical errors on the model histograms. The second is how does RooFit handle a weighted histogram for the model with respect to the errors. I have been reading up on how to handle these things. One reference that has proven helpful is the Barlow and Beeston paper in the Computer Physics Communications journal (Vol 77 Page 219-228). I have looked in the manual for RooFit and have found nothing speaking of problem one and a small footnote on problem two stating that it is in general incorrect to use the sum of the weights to equal the number of events. I am curious as to why this is incorrect? Can anybody comment on how these things are handled in RooFit and whether I can use RooFit to help me with my fits or whether I will need to construct my own tool to perform these fits.
I would prefer to perform an unbinned likelihood fit if possible. Thanks for any input on these questions.