This repository rasemble all the codes used to perform the fit of the BumpHunter test statistics distribution presented in the article : "Fit of the BumpHunter test statistic distribution and global p-valus evaluation".
The results were obtained using pyBumpHunter v0.4.0.
Link to pyBumpHunter repository
Link to the paper [not yet]
- Create a new python environement
python3 -m venv .env
source .env/bin/activate
- Install the required packages
pip install -r requirement.txt
- Run the test you want
The comand line comes with a number of arguments that are used to tune some parameters.
The available arguments are :- Nb : the number of bins in the scanned distribution (set to 20 by default)
- loc_path : Path to the directory where the final plots should be stored (default to 'results')
- tem_path : Path to the directory where the teporary files generated when scanning the PEs are stored (default to 'temp')
- width : Width of the window used by BumpHunter to scan the distribution (default to 1)
- bkg : Define the shapes of the background in the scanne distribution, can be either 'exp' or 'lin' (default to 'exp')
- Nbkg : Define the number of background event generated in the scanned histograms (default to 100k)
Example of command line used to run the test :
python3 fit_test.py --Nb 20 --bkg exp --Nbkg 1000 --width 1
Optionnally the script run_all.sh
can be used to run automatically all the tests to reproduce the results of the paper.
Warning : By default, running a test uses 10 CPU cores to parallelize the scans.
Make to that your machine can support it (or modify the number of workers at the line 291 of the code.)