Causal Inference & Causal ML
Go from 'X is correlated with Y' to 'X causes Y': potential outcomes, causal graphs, matching, natural experiments, and machine learning for treatment effects.
Before this: Machine Learning. Comfort with regression and basic probability from Python & Math for AI.
-
16 min
Step 1Causal inference in 15 minutes
Why correlation misleads and what causal questions look like.
Causal Inference - EXPLAINED!CodeEmporium
-
42 min
Step 2A brief introduction to causal inference
Brady Neal's course overview: confounding, interventions, and the big picture.
1 - A Brief Introduction to Causal Inference (Course Preview)Brady Neal - Causal Inference
-
8 min
Step 3Potential outcomes
The Rubin model: compare what happened with what would have happened.
The Rubin Causal model - an introductionBen Lambert
-
6 min
Step 4Causal graphs
Draw assumptions as arrows between variables.
Introduction to Causal GraphsLeslie Myint
-
12 min
Step 5Forks, chains, and colliders
The three graph patterns that decide what to control for.
Key Structures in Causal GraphsLeslie Myint
-
14 min
Step 6Causal effects from graphs
Backdoor and front-door criteria for handling unobserved confounders.
Causal Effects via DAGs | How to Handle Unobserved ConfoundersShaw Talebi
-
17 min
Step 7Propensity score matching
Compare treated and untreated units that were equally likely to be treated.
An intuitive introduction to Propensity Score MatchingDoug McKee
-
13 min
Step 8Instrumental variables
Use a variable that shifts treatment but not the outcome directly.
Introduction to Instrumental Variables (IV)Marginal Revolution University
-
33 min
Step 9Difference-in-differences
Compare changes over time between a treated and a control group.
9 - Difference-in-DifferencesBrady Neal - Causal Inference
-
24 min
Step 10Causal inference in Python
A conference talk on framing and estimating cause and effect with Python libraries.
An introduction to Causal Inference with Python – making accurate estimates of cause and effect fromPyCon AU
-
31 min
Step 11Machine learning for treatment effects
Estimate who benefits most from a treatment with EconML and CausalML.
Hajime Takeda - Introduction to Causal Inference with Machine Learning | SciPy 2024SciPy
py-why/EconML★ 4.9k -
22 min
Step 12Double machine learning
Adjust for many observed confounders with ML; still assumes none are unmeasured.
Double Machine Learning, Clearly Explained (Part 1)AI, ML, SWE, Tech Expert
DoubleML/doubleml-for-py★ 801 -
15 min
Step 13Causal discovery
Infer a plausible causal graph from observational data, with a Python example.
Causal Discovery | Inferring causality from observational dataShaw Talebi
Build it
Watching isn't learning. Pick one and put it on GitHub.
- Draw a causal graph for a question you care about, list which variables to adjust for, and defend each choice in a short README.
- Estimate the effect of job training on earnings with propensity score matching and compare it with a naive difference in means.
- Use EconML on a public marketing or uplift dataset to find which customer groups respond most to a treatment.