intro-to-causal-inference

A introduction to causal inference using common tools from the python data stack

Table of Contents

• Getting Started
  • Install graphviz
  • Clone the repository
  • Determine your installation preference
  • Install a new IPython kernelspec
  • Start up jupyter lab and open a notebook
• Acknowledgements
• Feedback

Getting Started

Install graphviz

You'll need graphviz for our first exercise notebook, to visualize causal graphs.

• Linux:
  • Depending on your distro, here are the possible commands
• MacOS:
  • You can easily install it via homebrew: brew install graphviz
• Windows:
  • The graphviz.org website has convenient installers for you

Clone the repository

In your terminal, use git to clone the repo to your machine.

git clone [email protected]:ronikobrosly/causal_inference_intro.git

If you are less comfortable with git, there is an easy alternative: You can simply download a zip file of it here :)

Determine your installation preference

Now that you've installed graphviz and cloned the repo locally, you'll need to ensure you have a working python environment set up. There are two ways you can go about doing this, shown as Options 1 and 2 below. Option 1 is the preferred method, as it is more likely to work across all platforms.

If you already have installed and are using Anaconda python on your laptop please use Option 2.

This tutorial requires python version 3.9 for the notebook exercises to properly work.

Option 1: installing via pip install in a virtualenv

Create a new virtual environment for this tutorial. You can do this a number of ways but one of the simpler involves the use of the virtualenv python package. If you are a Windows or Linux user, please use this guide. If you are a MacOS user you can follow this guide.

Name your environment causal_modeling

"Activate" this environment (see the above guide) and then run the following command in the root folder of this repo: pip install -r requirement.txt

This will install all the necessary packages for the tutorial.

As an optional step, you can try to run the check_environment.py file (in the root folder of the repo) while within your virtual environment. You can do so by running python check_environment.py in your terminal. It will alert you if you're missing any required python packages.

Option 2: installing via Anaconda python and the conda package manager

If you do not already have the Anaconda distribution of Python 3, please install it.

The next step depends on what platform you are using: 1) Mac and Linux vs 2) Windows

Mac and Linux:

You can then use the conda tool in your terminal to install the necessary packages via this command:

conda env create -f conda_env_mac.yml

"Activate" the new environment via:

conda activate causal_modeling

As an optional step, you can try to run the check_environment.py file (in the root folder of the repo) while within your virtual environment. You can do so by running python check_environment.py in your terminal. It will alert you if you're missing any required python packages.

Windows:

You can then use the conda tool in your terminal to install the necessary packages via this command:

conda env create -f conda_env_windows.yml

"Activate" the new environment via:

conda activate causal_modeling

As an optional step, you can try to run the check_environment.py file (in the root folder of the repo) while within your virtual environment. You can do so by running python check_environment.py in your terminal. It will alert you if you're missing any required python packages.

If the above Anaconda methods do not work:

You may have to create your own python 3.9.x environment using the conda command:

conda create --name causal_modeling python=3.9

"Activate" the new environment via:

conda activate causal_modeling

And then you'll pip install all dependencies within this environment

pip install -r requirements.txt

Install a new IPython kernelspec

Once the above is complete, you'll need to run the following commands:

python -m ipykernel install --user --name causal_modeling --display-name "Python (causal_modeling)"

Start up jupyter lab and open a notebook

In the terminal, execute jupyter lab.

Navigate to the notebooks directory and open your notebook of choice. You will probably be asked which environment you would like to use with the notebook. Select the Python (causal_modeling) environment you created in the step above.

Acknowledgements

I would like to like to acknowledgement the following individuals for creating public causal inference materials that were useful in the creation of this tutorial:

• Konstantinos Papadopoulos @ Analytics Mayhem
• Ramin Ghelichi @ Wayfair

Feedback

I love would to hear your feedback on these tutorial materials! Please send your comments to [email protected].