Binning in RooMCStudy (tutorial roostats/rs504)

Roofit and RooStats
1

Hi,

I’m using ROOT 5.25/02. I ran the roostats/rs500d macro to create a workspace with a gaussian signal over a flat background. This creates a model file which is the default input file for the roostats/rs504 macro. The rs504 macro runs some pseudo experiments and calculates the signal significance.

As the calculation was somewhat slow, I wanted to generate binned data which could help to speed things up. So I changed line 78 and 81 in rs504 and added the parameter “Binned(true)” to the argument list of the RooMCStudy constructor.

However, re-running the pseudo experiments with these changes yields a totally different average significance (0.78sigma instead of 29sigma).

What am I doing wrong?

Thanks and cheers,

Jochen

2

Hi Jochen,

Thanks for reporting that. That looks like a bug. I will run your macro shortly to confirm and will enter a Savannah bug report to have it fixed for the next release (next month).

Wouter

3

Hi,

this does not seem to be fixed in the latest stable ROOT version, v5-26-00b.

Please tell me what’s the plan for this.

Cheers,

Jochen

4

Hi Jochen,

I thought I had fixed this in the release. I’ll investigate and
get back to you.

Wouter