PELEE

Prerequisites

Conda Environment

(1) Install free conda from miniforge: https://github.com/conda-forge/miniforge

(2) follow the setup below:

conda create -n python3LEE python=3.7
conda activate python3LEE
conda install scipy pandas==1.0.5 matplotlib pyyaml tqdm scikit-learn jupyter
# Analysis currently uses iminuit 1.5.4 syntax
conda install -c conda-forge iminuit==1.5.4
pip install unitpy
conda install -c conda-forge uproot==3.11.6
conda install dask-core
conda install -c conda-forge xgboost==0.90
conda install -c conda-forge shap
pip install unitpy
conda install pytables

(the xgboost version needs to be specified to be compatible with the stored BDTs)

(3) To activate this local python setup type from the terminal:

conda activate python3LEE

Local Settings

You will need a file localSettings.py in the root directory of this repository that defines the root directory where all the data lives that contains:

main_path = "/Users/cerati/Notebooks/PELEE/"
ntuple_path = "/Users/cerati/Notebooks/PELEE/root_files/1013/"
pickle_path = "/Users/cerati/Notebooks/PELEE/pickles/"
plots_path = "/Users/cerati/Notebooks/PELEE/plots/"
dataframe_cache_path = "/Users/cerati/Notebooks/PELEE/cached_dataframes/

The path dataframe_cache_path is used to store loaded dataframes to disk for faster loading in the future when enable_cache=True is set for the data loading functions.

Unit Tests

Run unit tests to make sure core functionality is working as intended. These tests are also automatically run on GitHub for every PR to master. Since all tests are defined in source files that match the test_*.py pattern, they are automatically discovered and run by unittest.

python -m unittest discover

Tutorial

Run the MicroFit Tutorial.ipynb notebook to familiarize yourself with the MicroFit framework and how to run a basic two-hypothesis test.

Technote plots

The following scripts make the technote plots for each sub-analysis. These include the histograms of the signal and sideband channels, the covariance matrices (total and split by error source), tables with error budgets and constraint updates in TeX format, chi-square distributions, two-hypothesis tests, and the FC corrected Asimov sensitivity scan. These scripts can take up to 2-3 hours to run if they are run from scratch without already existing cached dataframes.

Technote plots for the new signal analysis

All technote plots for the analysis with the new signal model can be produced by running

python new_signal_ana_technote_plots.py

while inside the scripts directory. Make sure to make a directory called full_ana_remerged_crt_output before running the script. The script is fully automated and loads all necessary data and produces the detector variations it needs. All plots that are produced will be placed into full_ana_remerged_crt_output.

Technote plots for the old signal analysis

All technote plots for the analysis with the old signal model can be produced by running

python old_signal_ana_technote_plots.py

while inside the scripts directory. Make sure to make a directory called old_model_ana_remerged_crt_output before running the script. The script is fully automated and loads all necessary data and produces the detector variations it needs. All plots that are produced will be placed into old_model_ana_remerged_crt_output.

ubneutrinos / PELEE

readme