RooFit: Change of behavior of ProjWData() with 5.32

Roofit and RooStats
#1

Dear RooFit experts,

With the change of the root version to 5.32 I noticed a change in the behavior of the RooCmdArg ProjWData(), when plotting a self written pdf containing event-by-event variables. The fitting still works fine, using ConditionalObservables() with a RooDataSet, but when it comes to plotting (using ProjWData() and the same RooDataSet) something leads to strange values of the event-by-event variable, when the evaluate() function is called. It seems to me, that with the new RooFit version, another way of handling the event-by-event variables was introduced. Some kind of averaging or different way of projecting the variables.
Is there any way to reproduce the 5.30-way of plotting?
I’m sorry, that I cannot be more specific, but to see through the whole plotting procedure is not that easy.

Best regards

#2

Hi,

There was no intential change of behavior, so this might be the result of a bug.
Some code related to handling of conditional observabes was fixed in patch
release 5.32.01. Have you tried that version already (or the .02 version that is out
meanwhile).

If the patch version doesn’t help, could you post a simple example that reproduces the problem?

Wouter