an example analysis code with alphatwirl and data frames
- the example has two steps
- create data frames from ROOT Trees with alphatwirl
- This step creates input data frames to the 2nd step. The design of the data frames depends on what you do in the 2nd step.
- analyze data frames with pandas in python
- This step is the main part of the analysis.
- The example highlightes data frame operations.
- general: the code doesn’t know the names or the values of the categorical variables, or even the number of the dimensions of the categorical variables. That means that the same code will just work even if you add a new dimension in the categorization.
- concise: easy to write vectorized code. never need to write “for” loops to loop over the rows of data frames. rarely need “if” statements. You won’t end up writing nested “for” loops and “if” statements, which is error-prone, hard to understand and maintain
- create data frames from ROOT Trees with alphatwirl
Log in to soolin.
Move to a work dir:
cd /some/work/dir/The code runs in any envriomment with Python 2.7, ROOT 6, and several other common libraries.
In this example, we will run in cmsenv.
Source cmsset_default.sh if it is not done yet.
source /cvmfs/cms.cern.ch/cmsset_default.shCheck out cmsenv if it is not done yet.
export SCRAM_ARCH=slc6_amd64_gcc630
cmsrel CMSSW_9_4_3Enter the cmsenv:
cd CMSSW_9_4_3/src/
cmsenv
cd ../../Check out this repo:
git clone --recursive git@github.com:TaiSakuma/xi-roc-cafe.gitRead trees for QCD
./xi-roc-cafe/twirl_mktbl_heppy.py --components QCD_HT100to200_madgraphMLM QCD_HT200to300_madgraphMLM QCD_HT300to500_madgraphMLM QCD_HT500to700_madgraphMLM QCD_HT700to1000_madgraphMLM QCD_HT1000to1500_madgraphMLM QCD_HT1500to2000_madgraphMLM QCD_HT2000toInf_madgraphMLM -i /hdfs/SUSY/RA1/80X/MC/20161021_B03/ROC_MC_SM -o ./tbl_001/QCD --parallel-mode htcondor --max-events-per-process 500000 --logging-level INFORead trees for T1tttt
./xi-roc-cafe/twirl_mktbl_heppy.py --components SMS_T1tttt_madgraphMLM -i /hdfs/SUSY/RA1/80X/MC/20161021_B04/ROC_MC_SMS -o ./tbl_001/T1tttt --parallel-mode htcondor --max-events-per-process 500000 --logging-level INFO --susy-smsThe above two commands create a set of data frames in the directory tbl_001/.
They also create a temporary dir _ccsp_temp/. Remove it if you don't need:
rm -rf _ccsp_temp/The same data frames are in example_tbl/tbl_twirl in this repository.
Instead of watining for the jobs to finish, you can copy them and continue the example.
cp -a ./xi-roc-cafe/example_tbl/tbl_twirl ./tbl_002/Copy a table with the relations between data sets and processes:
cp -a xi-roc-cafe/tbl/tbl_cfg_component_phasespace_process.txt tbl_002/QCD/Analyze the data frames:
./xi-roc-cafe/tbl_roc.py --dir ./tbl_002/This will print the ROC for two variables xi and minOmegaTilde against QCD and two mass points of T1tttt in bins of HT and njets.