Code and python config files for ttH, H -> tautau analysis with matrix element techniques

Since FUSE appears to be buggy, accessing files via /hdfs mountpoint is really slow (especially when a given folder contains thousands of files). One of the (viable) solutions is to build the HDFS plugin ourselves because CMSSW stack doesn't ship ROOT with HDFS support. This can be done by executing the following command which builds the plugin and sets necessary environment variables to access any file via the HDFS protocol:

source $CMSSW_BASE/src/tthAnalysis/HiggsToTauTau/misc/set_env.sh

The files residing on /hdfs can be read via

TFile * f = TFile::Open("hdfs:///your/desired/path");

anywhere. NB! The plugin does not support writing via this protocol.

Clone this repository into $CMSSW_BASE/src/tthAnalysis/HiggsToTauTau, i.e.

git clone git@github.com:HEP-KBFI/tth-htt.git $CMSSW_BASE/src/tthAnalysis/HiggsToTauTau

Set up your CMSSW working environment, do scram b -j8 in $CMSSW_BASE/src.

One needs to install ClassicSVfit and its dependency SVfitTF in order to compile the whole project; do

git clone https://github.com/veelken/ClassicSVfit TauAnalysis/ClassicSVfit
git clone https://github.com/veelken/SVfitTF TauAnalysis/SVfitTF

and compile the project..

We currently have implemented the analysis for 7 channels (plus charge misid and MVA), the runtime configuration files of which are test/tthAnalyzeRun*.py.

Review the analysis section in one of these files, e.g. test/tthAnalyzeRun_3l_1tau.py:

version = "2016Jul19_dR03mvaTight"

analysis = analyzeConfig_3l_1tau(
    outputDir = os.path.join("/home", getpass.getuser(), "ttHAnalysis", version),
    executable_analyze = "analyze_3l_1tau",
    lepton_selections = [ "Tight" ],
    hadTau_selection = "dR03mvaTight",
    central_or_shifts = [ "central"],
    max_files_per_job = 30,
    use_lumi = True, lumi = LUMI,
    debug = False,
    running_method = "sbatch",
    num_parallel_jobs = 4,
    histograms_to_fit = [ "EventCounter", "numJets", "mvaDiscr_3l", "mTauTauVis" ],
    select_rle_output = True)

Go to the test folder and run the file:

cd $CMSSW_BASE/src/tthAnalysis/HiggsToTauTau/test
python tthAnalyzeRun_3l_1tau.py

and wait until you're prompted to run the job exectution, e.g.

Start jobs ? [Y/n]

Press enter and wait til the script tells you "Done". Or, you could decline and run make later:

make -f /home/$USER/ttHAnalysis/2016Jul19_dR03mvaTight/Makefile_3l_1tau -j 4

This runs a python script (sbatch_analyze_3l_1tau.py) which submits the analysis jobs to SLURM and waits them to complete (approx 30 min per channel). All the results are stored in outputDir defined in tthAnalyzeRun_3l_1tau.py. The datacards are located under datacards; the run-lumi-event numbers of selected events are under output_rle`.

1. mode

   1.a) VHbb -- runs the analysis on all the samples, except for the FastSim samples; the final products of this analysis are datacard files that contain the distributions of various variables used in signal extraction across all event categories (signal, ttbar, ewk, fakes, rares etc); these datacards are used as an input to combine which calculates the limits;

   1.b) forBDTtraining -- runs the analysis exclusively on the FastSim samples; the final products are root files which contain an Ntuple; these Ntuples are used as an input to a BDT training algorithm; no datacards, no limits here;

2. use_prod_ntuples

   2.a) disabled -- runs the analysis on ,,VHbb Ntuples'', i.e. the original Ntuples which we produced ~1 year ago with Heppy; these Ntuples contain all information;

   2.b) enabled -- runs the analysis on ,,production Ntuples'', i.e. the Ntuples that are derived from the ,,VHbb Ntuples'' by removing branches which we don't use in our analysis; by default, the events are required to pass a very loose preselection containing at least 1 fakeable lepton and 1 fakeable tau passing the medium tau WP; in case of FastSim samples, there is no preselection applied, only the unused branches are stripped.

If you want to produce the ntuples for the synchronization exercise, run

sync_ntuples $CMSSW_BASE/src/tthAnalysis/HiggsToTauTau/test/sync_ntuples_cfg.py

Check the input/output parameters in the configuration file first, though. The resulting sync Ntuple can be compared against the other Ntuples w/ appropriate macros.