Pdf projection difference between root 6.36.06 and root 6.38.00

Roofit and RooStats
1

Hello,

I am trying to fit a model to some data, with 3 regions, and then draw the projections. In root 6.36.06, the plots look fine. In 6.38.00, the normalisation of the pdf seem crazy in the projection (the fits in both version return identical numbers). How should I do these projections in root 6.38 to recover the proper behavior ?

I attached a simple macro to illustrate the problem, and two plots. I simply do :

root.exe miniProj.C

6.36.06 was installed on my mac with conda, and I compiled 6.38.00 from the source (but it was checked that the issue is also there in a more official installation).

Best..

Jean-Baptiste

miniProj.C (2.0 KB)

r63800.pdf (31.7 KB)

r63606.pdf (32.2 KB)

2

Maybe the argument Normalization(nEvents, RooAbsReal::NumEvent) helps as suggested in RooFit normalization problem on release v6.38/00 · Issue #20827 · root-project/root · GitHub

fyi @jonas

3

Hello,

Thanks a lot for the hint. If I do this, nEvents should be equal to the total number of events in the dataset (i.e. not the number of events expected in the particular category which is projected).

Is this expected that in 6.38, we need to do this explicit normalisation ?

Best,

Jean-Baptiste