This code is set up to study the model-dependence in H->gamma gamma fiducial differential cross sections.
We are training a BDT to parametrize the response matrix. The response matrix is the object that maps events from particle 
level to detector level. The training set consists of events comming from the four dominant Higgs production mechanisms
at the LHC, that are ggF, VBF, VH and ttH
Closure plots compare the number of events in each detector level bin predicted by the BDT and compare them to the truth
inferred from the MC test set. It is important to not test the BDT on any events used for the training.
The effect of using different response matrices is studied and the unfolded particle level spectra are plotted.

Important Ipython notebooks are:
Train_Configurable_Refactor.ipynb
ScriptForResponseMatrixAndClosurePlots.ipynb
EffectOfDifferentResponses.ipynb
DiffXsecHistos.ipynb

Important python files are:
angles.py
train.py
util.py
chi2_stats.py
LikelihoodProfile.py
plotting.py

A DESCRIPTION OF THE NOTEBOOKS' CONTENT IS GIVEN IN THE FOLLOWING:

1) Train_Configurable_Refactor.ipynb
in this notebook the BDTs, classifiers, are trained. The already trained classifiers are stored in:
/mnt/t3nfs01/data01/shome/jandrejk/higgs_model_dep/MoriondAnalysis/classifiers/
and are available until 11.11.2017!
Each classifier has a .pkl, .root and a .txt file associated.
The .txt file contains details about the training and the hyper-parameters.

2) ScriptForResponseMatrixAndClosurePlots.ipynb
This notebook loads trained classifiers. It produces the closure plots and saves the histograms of the
closure plots in form of matrices in .npy files. These files are used in the notebooks 3) and 4) to 
compute the response matrices without the need of loading the classifiers from scratch.

3) EffectOfDifferentResponses.ipynb
This notebook likelihood plots the profiles around the best fit fiducial differential cross sections 
saves them (if a path is provided).
The profiles are computed with the BDT predicted response matrix and with the matrix build from the SM MC  
test sample. The response matrices are loaded from the directory specified in the notebook 
"ScriptForResponseMatrixAndClosurePlots". 

4) DiffXsecHistos.ipynb
This notebook is plotting the distribution of fiducial differential cross sections for different models 
and 2 scenrios:
1) The response matrix is the one coming from the BDT prediction 
2) The response matrix is the one coming from the SM MC test sample
The differential x-sec are computed with respect to recoPt and recoNjets2p5. For that the profiles around 
the best fit fid diff cross sections are computed in the same way as in the notebook "EffectOfDifferentResponses" 
but this time the profiles are plotted together in a histogram.