RooFit: non-integer bin contents

Roofit and RooStats
1

Is it possible for RooFit to fit histograms with non-integer bin contents. I realise that RooFit was conceived to predict numbers of counts and fit them to data which is also in numbers of counts, but I’d like to use the object orientated modeling functionality to fit a graph with non-integer (and in fact some negative) bins. RooFit seems to round all the input to the nearest integer and set negative bins to zero. Other than multiplying all the bins by some large number so that the non integer parts can be neglected is there any way to get around this?

Cheers

Hugh

2

Hi Hugh,

This happens in the default mode of visualization of a dataset because Poisson errors can only be calculated for integer bin counts. You can choose to show your dataset with sum-of-weight errors instead, which can also show non-integer bin content:

data->plotOn(frame,DataError(RooAbsData::SumW2)) ;

Wouter