Jupyter Notebooks

This page contains a collection of free Jupyter notebooks illustrating specific causal inference concepts and considerations. Each notebook focuses on a particular question, misconception, or modelling behaviour, and is shared in both .ipynb and PDF format.

  • Free

When Biased CATE Estimates Are Still Useful

  • Download
  • 2 files

Demonstrates that biased CATE estimates can sometimes still be useful for ranking treatment effects, even when effect magnitudes are incorrect.

  • Free

SHAP Values Shouldn't Be Interpreted Causally

  • Download
  • 2 files

Demonstrates that SHAP values reflect predictive importance rather than causal importance.

  • Free

Adjusting For Variables Correlated With Mediator

  • Download
  • 2 files

Explores whether adjusting for variables correlated with a mediator necessarily introduces bias.

  • Free

Granger Causality is not True Causality

  • Download
  • 2 files

Demonstrates why Granger causality does not imply true causal effects.

  • Free

On the Different Sensitivity of ATEs and CATEs to Model Misspecification

  • Download
  • 2 files

Explores the different sensitivity of ATE and CATE estimates to model misspecification in experimental settings.