Causal Inference & Experiments

Read these first

1. 01Difference-in-Differences

   A way to measure the effect of a policy or event by comparing how much a treated group changed before-to-after against how much an untreated group changed over the same window, so that anything moving both groups equally cancels out.

2. 02Directed Acyclic Graphs for Causality

   A way to draw what causes what before you run a single regression, so you can tell in advance which variables to control for and which ones will wreck your estimate if you touch them.

3. 03Propensity Score Methods

   A way to compare a treated group against an untreated group as if treatment had been handed out at random, by first summarizing every unit's covariates into a single number, its odds of being treated, and matching or weighting on that number instead of on the raw covariates.

4. 04Regression Discontinuity Design

   A way to estimate a causal effect by exploiting a sharp, arbitrary cutoff rule, comparing units that barely qualified against units that barely missed, since near the threshold that's effectively a coin flip.

5. 05The Potential Outcomes Framework

   The formal way to define what a causal effect even means, the difference between what would have happened to the same unit with and without a treatment, and why that definition makes clear exactly why correlation isn't causation.

Then the rest

1. 06ATE, ATT and Other Treatment Effect Estimands
2. 07Bad Controls and Post-Treatment Bias
3. 08Colliders and Collider Bias
4. 09Confounding and the Backdoor Criterion
5. 10Doubly Robust Estimation
6. 11Event Study Designs
7. 12Experimental Design: Blocking and Factorial Designs
8. 13Ignorability and Unconfoundedness
9. 14Inverse Probability Weighting
10. 15Matching and Nearest-Neighbour Estimators
11. 16Mediation Analysis
12. 17Parallel Trends and Pre-Trend Tests
13. 18Randomization Inference in Experiments
14. 19Simpson's Paradox
15. 20Structural Causal Models and Counterfactuals
16. 21The Exclusion Restriction and Instrument Validity
17. 22The Local Average Treatment Effect
18. 23The Synthetic Control Method
19. 24Causal Discovery from Observational Data
20. 25Heterogeneous Treatment Effects and Causal Forests
21. 26Sensitivity Analysis for Unobserved Confounding