# README file for HiggsAnalysisTools

How to get the code
-------------------
cvs co -d HiggsAnalysisTools UserCode/emanuele/HiggsAnalysisTools
cvs co -d EgammaAnalysisTools UserCode/emanuele/EgammaAnalysisTools
cvs co -d CommonTools UserCode/emanuele/CommonTools
cd HiggsAnalysisTools/config
cvs co -d kfactors_Std HiggsAnalysis/HiggsToWW2Leptons/data/kfactors_Std
cd ..

How to choose the selection that is compiled
--------------------------------------------
The main() function is in the file src/HiggsApp.C
Which one of the selections are compiled and run depends on the variable: Application.
To change it modify: include/Application.hh 

the code containing the H->WW->2l2nu selection is:
- src/HiggsMLSelection.cc
- src/CutBasedHiggsSelector.cc

How to compile and run the code
-------------------------------
cp /afs/cern.ch/user/e/emanuele/public/4Latinos/pdfs_MC.root .
make -j4 HiggsApp

./HiggsApp <listwithfiles> <output-prefix> <ismc>

- run on MC example: 
./HiggsApp testH160_Fall11.txt Hww160 1

where testH160_Fall11.txt is the list of files (vecbos ntuples)

- run on data example (JSON is in config/json/goodCollisions2011.json):

edit the file config/higgs/2l2nuSwitches.txt and modify the first two lines to be:
isData                  1
goodRunLS               1

./HiggsApp doubleelectron.txt prefix 0

this will run the standard selection, and produce 4 selected trees (one for each channel: ee/mumu/emu/mue):
prefix-datasetMM/EE/EM/ME.root

these are the 'unweighted' latinos trees.


How to run in batch:
-------------------
The output is typically big, so typically one runs in batch and each job copies the output
on a computer disk through ssh at the end of running. 
For each 'dataset', i.e. txt list file in cmst3_42X/, the cmst3_submit_manyfilesperjob.py has to be executed.

Edit the script cmst3_submit_manyfilesperjob.py to change the computer where you want the ouptut. Change the three lines specyfying it.

diskoutputdir="/dirthatyoulike..."
    os.system("ssh -o BatchMode=yes -o StrictHostKeyChecking=no nameofthepcthatyoulike mkdir -p "+diskoutputmain)
    outputfile.write('ls *.root | grep -v Z_calibFall08 | xargs -i scp -o BatchMode=yes -o StrictHostKeyChecking=no {} nameofthepcthatyoulike:'+diskoutputmain+'/{}\n') 

- run on all MC samples:
./submitMassDependentAnalysis.pl -p MC_Submission_V1
will create the directory  MC_Submission_V1 with cfgs and log files in that directory, 
and another  MC_Submission_V1 on the pc where it will write the output ROOT files

- run on all data samples:
./submitMassDependentAnalysis_data.pl -P Data_Submission_V1


How to weight / merge latinos trees after the batch jobs are done:
------------------------------------------------------------------
cd macro/
ln -s dirwheretheMCfilesare results
ln -s dirwheretheDatafilesare results_data

chmod a+x fromSingleJobResToRooDataSet.sh
chmod a+x fromSingleJobResToRooDataSetData.sh 

- MC:
./fromSingleJobResToRooDataSet.sh 1000 (takes a lot of time!)
will compute and store all the important weights in the latinos trees (for 1000 pb-1).

The important weights are:
"baseW": weight / 1fb-1
"puW": to reweight for the PU profile of te Full2011 data taking
"effW": data / MC scale factors
"kfW": k-factor (different from 1 only for Higgs and DrellYan)

- data:
./fromSingleJobResToRooDataSetData.sh
will do the same to the data


Where the important files are:
------------------------------

- MC: 
results/datasets_trees:  contains the weighted trees
results/datasets_trees_skim: the same trees, but only with events passing the HWW pre-selection (WWSel OR WWSel1j bits)

- data:
results_data/datasets_trees
results_data/datasets_trees_skim