FNAL LPC CMS Data Analysis School (CMSDAS): Statistics Short Exercise

Introduction

This is a set of tutorials for the CMS Data Analysis School (CMSDAS) Statistics Short Exercise that are implemented in Jupyter notebooks. The tutorials provide a concise introduction to key ideas in statistics and a few of the statistical analysis tools commonly used in high-energy physics.

Notebooks

notebooks	description
0/exercise_0.ipynb	a very short PyROOT/RooFit tutorial
1/exercise_1a.ipynb	analyzing a single count using RooFit/RooStats
1/exercise_1b.ipynb	analyzing a single count using RooFit/RooStats
2/exercise_2.ipynb	analyzing three counts using RooFit/RooStats
3/exercise_3a.ipynb	analyzing three counts using combine
3/exercise_3b.ipynb	a realistic counting experiment using combine
3/exercise_3c.ipynb	systematic uncertainties for data-driven background estimates using combine
4/exercise_4.ipynb	binned fit to Run I H->gg data (optional)
5/exercise_5.ipynb	unbinned fit to Type Ia supernovae distance modulus/red shift data (optional)
6/exercise_6.ipynb	histogram template analysis using combine

Setup Instructions

1. Go to https://coffea.casa and log in with your CMS credentials.

2. Start a new server with the image named Combine / Python 3.10 / ROOT 6.26.4.

3. In the terminal window, download the tutorials by typing:

   git clone https://github.com/FNALLPC/statistics-das.git

   Make sure to use https and NOT ssh on coffea.casa.

4. Initialize the environment:

   conda init
   source  /home/cms-jovyan/.bashrc

5. To open a notebook, e.g. statistics-das/0/exercise_0.ipynb, simply click on the file.

6. When using combine or text2workspace.py you may encounter "command not found" error. If so, open a terminal and update the following environment variables:

   export PATH=$PATH:/tmp/HiggsAnalysis/CombinedLimit/build/bin/
   
   export LD\_LIBRARY\_PATH=$LD\_LIBRARY\_PATH:/tmp/HiggsAnalysis/CombinedLimit/build/lib/
   
   export PYTHONPATH=/tmp/HiggsAnalysis/CombinedLimit/build/lib/python
   
   

7. Have a coffee.

Notebook Tips

Jupyter notebooks are structured into cells. It is recommended that you execute the notebooks one cell at a time to ensure that each cell works as expected. And, yes, cells should be executed sequentially! Of course, if you're careful, you can go back and re-execute cells. But remember jupyter notebooks are ``sticky'' in that the notebook pays attention to the order in which cells have been executed. This can lead to confusion. If things get really messed up, it's usually better to re-launch the jupyter kernel and start again from scratch.

Some useful shortcuts

• Use esc r to disable a cell
• Use esc y to reactivate it
• Use esc m to go to markdown mode. Markdown is the typesetting language used in jupyter notebooks. In a markdown cell, double tap the mouse or glide pad (on your laptop) to go to edit mode.
• Shift + return to execute a cell (including markdown cells).
• If the equations don't typeset, try double tapping the cell again, and re-execute it.

Links & Resources

• Exercise repository on github

• 2026 LPC CMS DAS lecture slides

• Legendary 2023 LPC DAS lecture slides

• Javier Duarte's lecture slides

• Nick Wadle's 2022 HCP Summer School slides