Tag And Probe 2019

Introduction

The Tag and Probe method consists in calculating the efficiency of the data itself without the dependency on the simulations, and with a small or easily subtracted, fake rate [2]. For that, this method uses known resonances (e.g J/$\psi$, Z, $\Upsilon$) to select particle of a desired type and probe the efficiency of this particle in a determined selection criteria.

Tags and Probes objects

In the Tag and Probe method, there are two objects:

*Tag

It is an object that passes a set of very tight selection criteria designed to isolate the required particle type (usually an electron or muon). And the fake rate for passing tag selection criteria should be very small ($<1\%$).

*Probe

It is a generic set of the desired particle type (with very loose selection criteria). It is selected by pairing these objects with tags, such as the invariant mass of the combination that must be consistent with the mass of the resonance.

To eliminate the background, one can use any of a variety of background subtraction methods such as fitting.

Tight and Loose Criteria

As tight criteria for the Tags, we use isTightMuon() function which has many functions that can determine a muon with very small fake rate. The implementation may be seen in this page

We choose Probes, by a loose criteria, only requiring them to be in the MuonCollection. We can calculate the invariant mass by pairing those two and see if it is consistent with a resonance. In the code, we use the *TLorentzvector*, a ROOT Class. We compute some kinematics variables like pT (Transverse Momentum), $\eta$(Pseudo-Rapidity) and $\Phi$ and extract what we need and record the values in a ROOT tree.

Efficiencies

The definition of the probe object depends on the specification of the selection criterion being examined. The efficiency itself is measured by counting the number of “probe” particles that pass the desired selection criteria:

For muons the total efficiency measurement can be broken down (factorized) into sequential measurements. The steps can be, for example: muon tracking efficiency muon tracking efficiency, stand-alone muon reconstruction efficiency, identification efficiency, isolation efficiency and the online or total trigger efficiency. The total efficiency is given by the product:

The estimate of this total efficiency is important, with the luminosity, to calculate the cross section of a determined process:

In this first work, we will focus on the first step of the process: extracting the sample for the efficiency computation. The workflow representing this first step can be seen in the figure \ref{fig:Tag_Probe}.

The structure

To access the data you must use the CMSSW, the structure that have the analysis environment. To access the data you must use the CMSSW, the structure that have the analysis environment. To do so, you have two ways:

*Getting the CMSSW VM Image

You must download the VM Virtual Box. Make sure that the version you downloaded is working with your CMS VM Image. You can read more [[http://opendata-dev.web.cern.ch/docs/cms-virtual-machine-2011][here]. You should use the CMS SHELL in the CMS Open Data VM 1.5.1 version as you can see [[http://opendata-dev.web.cern.ch/docs/cms-getting-started-2011][here]. With other versions (CMS-OpenData-1.2.0" or "CMS-OpenData-1.3.0) you can use the normal terminal. If you use the terminal they cannot construct because the SL7 is not compatible with this CMSSW version

*Using Docker Container

If you are having problem with virtual machine you can use the Docker Container. The instruction of installation are in this page. As you can see in the page, you can fetch a CMSSW image, and create and start a container using the docker run command docker run --name opendata -it clelange/cmssw:5_3_32 /bin/bash, but with that you will not be able to use ROOT. To do so, you must use the command:

docker run --name opendata -it clelange/cmssw:5_3_32 /bin/bash

but with that you will not be able to use ROOT. To do so, you must use the command:

sudo docker run --name opendata -it -e DISPLAY=$DISPLAY -v /tmp/.X11-unix:/tmp/.X11-unix clelange/cmssw:5_3_32 /bin/bash

and you will able to use ROOT.

To list the containers you created use the command:

sudo docker container ls -a

To open a container, just use the command with the ID of the container that tou can see in hte list of the containers you crearted.

sudo docker attach <ID_Container>
sudo docker attach <ID_Container>

Now, lets create the CMSSW structure. In this example we are using the CMSSW_5_3_32. After you emulate you CMS Open Data ova with success, you must created a CMSSW environment and change to the CMSSW_5_3_32/src/ directory (Notice that you use the container you do not need to use the the commands below):

cmsrel CMSSW_5_3_32
cd CMSSW_5_3_32/src/

Launch the CMS analysis environment.

cmsenv

Create a working directory for the analyzer and go to that directory

mkdir TagAndProbeSample

At this point, you must copy this Git to your area using the command:

git clone -b master git@github.com:cms-legacydata-analyses/TagAndProbe.git

Go to this directory that you copied from Git and compile the code:

cd TagAndProbe
scram b

After that go to test folder and run the python file:

cmsRun TagAndProbe_Model_Trigger.py

Wait and you will see a root file, named TagAndProbe2019.root, that will be created. You can open this file and see its contents:

root -l TagAndProbe2019.root
new TBrownser

We use the J/$\psi$ resonance in this example and make an output tree (root file) with the same content as the CMSSW TagAndProbe method. If you want study another resonance, just open the config file (TagAndProbe _Model_Trigger.py) and modify the range of mass in the line 43 and 44:

minResonanceMass = cms.double(2.8),# in GeV
maxResonanceMass = cms.double(3.4)# in GeV

You can change the dataset. Just change the address of the root file of the new dataset as you can see below:

'root://eospublic.cern.ch//eos/opendata/cms/Run2011A/MuOnia/AOD/12Oct2013-v1/20000/0045B338-934C-E311-BEF1-0026189438B0.root'     

and the triggers in this line:

PathName = cms.untracked.vstring(["HLT_Dimuon10_Jpsi_Barrel_v","HLT_Dimuon7_PsiPrime_v"]), 

Datasets and MC

The dataset used is here

A MC sample can be found in this page

Analysis notes

CMS-AN 2009/111 TP Note

CMS-EWK-10-005

Other Twikis

TagAndProbe