This is the code used to perform fitting, searching and limit setting for the Dijet+ISR analysis. The following people can help you if you have issues :

• Yvonne Ng - ying.wun.yvonne.ng@cern.ch (running and debugging)
• Kate Pachal - Katherine.Pachal@cern.ch (conceptual fitting issues)
• Sam Meehan - samuel.meehan@cern.ch (docker issues)

These are some of the key external references for this package

• ATL-COM-GEN-2017-001 : How to run the code
• 1101.0390 : Original BumpHunter paper
• CERN-THESIS-2015-179 : Kate's thesis

This code relies on the AnalysisBase release of ATLAS and there are two ways to allow for this use.

• CVMFS : Working on lxplus (or your local Tier3) which has access to CVMFS_
• Docker : Working with a docker image of AnalysisBase The interaction with GitLab to obtain the code, compile, and run it are largely the same, with the primary difference being the initial setup. Start by creating a high level directory in which your work will live

mkdir Fitting
cd Fitting

and proceed to setting up the environment with which you will work.

Set up the AnalysisBase release (setup_cvmfs.sh script)

setupATLAS
asetup AnalysisBase,21.2.3,here
export PYTHONPATH=./install/InstallArea/x86_64-slc6-gcc62-opt/src/Bayesian/plotting/PythonModules:${PYTHONPATH}

This describes the initial setup steps to be able to use a docker image for working with this code

• Start by installing the docker application on your local machine.
• Obtain the docker image for AnalysisBase : docker pull atlas/analysisbase:21.2.3
• Launch this docker image from the appropriate location : docker run --rm -it -v $PWD:$PWD -u root -e DISPLAY -w $PWD atlas/analysisbase:21.2.3
  • There are a number of arguments here and you can parse them here (documentation).
    However, the main thing to appreciate is that these will "bind" the local directories into the docker image so that you can access local files (code, text, etc.) on your laptop using whatever editor you like and have access to run and build it within the image
  • It is necessary to launch this from a sufficienctly top level directory such that the subdiretories with the code and the input histograms for fitting are contained beneath it
• Setup the work environment : source /home/atlas/release_setup.sh
  • It tells you to do this when you launch the image
  • This is equivalent to setupATLAS & asetup AnalysisBase,21.2.3,here in CVMFS but the output is not the same

Start by obtaining the project in the normal way

git clone https://gitlab.cern.ch/ywng/r21StatisticalAnalysis

Using cmake to configure, build, and install the code :

mkdir build
cd build

# configure
cmake ../source
# building
make 
# installing
make install DESTDIR=../install

The final step is to export the executables from the building via the setup.sh script in your install area

source ../install/InstallArea/${AnalysisBase_PLATFORM}/setup.sh

The code is now set up and compiled and ready to fit some spectra.

After you build the code the first time, the setup of the work area will be largely the same in terms of configuring the software release. If you are only going to rerun the fitting then you only need to execute the source ../install/InstallArea/${AnalysisBase_PLATFORM}/setup.sh setup command. However, if you modify the fitting, BumpHunter, or limit setting, then it will (of course) be necessary to recompile the package using cmake.

The fitting inputs are generally stored on eos at /eos/atlas/atlascerngroupdisk/phys-exotics/jdm/dijet/statsinputs/RunII/.

One example of the inputs is here :

/eos/atlas/atlascerngroupdisk/phys-exotics/jdm/dijet/statsinputs/RunII/ICHEP_DijetISR/inputs/hist_20160801/OUT_dijetgamma_mc/datalike-noNeff/

These inputs can be stored anywhere as the path will be specified in the [LINK TO RUN SCRIPTS] script which is the top level script you will run.

A comprehensive guide to the technical and pedagogical aspects of running the code if found in the references above.
However, a bare minimum start to running this is to start with the SearchPhase which itself will perform the fitting and execute the BumpHunter algorithm. This can be run on its own by creating a config file similar to the ones found in the configurations directory. However, there exist steering scripts such as Run_gjjSearchPhase.py which facilitates running this on batch systems, among other things.

To try a single first run

• Obtain the inputs /eos/atlas/atlascerngroupdisk/phys-exotics/jdm/dijet/statsinputs/RunII/ICHEP_DijetISR/inputs/hist_20160801/OUT_dijetgamma_mc/datalike-noNeff/
• Within the Run_gjjSearchPhase.py steering scriptconfigure the user input parameters to point at the contents of this input file
• Execute the script as python Run_gjjSearchPhase.py

You should see the initial fit performed and then a number of pseudoexperiments for estimating the error and finally the execution of more pseudoexperiments to estimate the significance of the largest excess. For details beyond this, please refer to ATL-COM-GEN-2017-001