Details
-
Bug
-
Status: Closed (View Workflow)
-
High
-
Resolution: Fixed
-
None
-
None
-
All
Description
When defining a model with a Poisson constraint with a global observable, the corresponding Poisson constraints is evaluated with noRounding=false, which means that the global observable value from the Asimov data set is treated as an integer, resulting in a wrong estimation of CLb and expected limits.
This is is not the case when defining the global observable as a real observable, because in that case the Poisson term has the flag noRounding=true.
This has been reported in
https://root-forum.cern.ch/t/roostats-hypothesis-inversion-demo-on-off-problem-how-to-handle-observables/40318/
The attached macros also show the problem.
The attached macro show a