Foundations
Routing Summary
This folder covers the conceptual foundations of applied econometrics and causal inference. Contains 4 notes plus the Randomization Inference sub-topic.
- Need the four FAQs framework for empirical research? → Research Questions in Econometrics
- Need why randomization is the gold standard? → The Experimental Ideal
- Need potential outcomes and selection bias? → The Selection Problem
- Need DAGs, d-separation, backdoor adjustment, valid adjustment sets? → Directed Acyclic Graphs
- Need Fisher randomization tests, sharp vs weak nulls, studentized/permutation inference? → Randomization Inference
Sub-topics
| Sub-topic | Notes | Covers |
|---|---|---|
| Randomization Inference | 5 | Fisher randomization test & the sharp null, sharp vs weak (Neyman) nulls, studentized randomization tests with dual finite-sample/asymptotic validity, permutation tests & exact inference — Wu & Ding (2021) |
Concept Map
| Concept | Note | Type | Depends On | Key Result |
|---|---|---|---|---|
| Four FAQs framework for organizing research | Research Questions in Econometrics | concept | Mostly Harmless Econometrics - Overview, The Experimental Ideal, The Selection Problem | Good research answers a causal or descriptive question clearly |
| Random assignment as causal inference benchmark | The Experimental Ideal | concept | The Selection Problem, Research Questions in Econometrics, Regression and the CEF | Randomization eliminates selection bias by design |
| Potential outcomes, selection bias decomposition | The Selection Problem | concept | Research Questions in Econometrics, Mostly Harmless Econometrics - Overview | Selection bias = E[Y0i given Di=1] - E[Y0i given Di=0] |
| DAGs, forks/chains/colliders, backdoor adjustment, d-separation | Directed Acyclic Graphs | concept | The Selection Problem, The Experimental Ideal | Conditioning on valid adjustment set (no colliders, all forks/chains blocked) recovers causal effect from observational data |
Notes
- Research Questions in Econometrics — CONTAINS: Four FAQs framework, causal vs descriptive questions, organizing empirical research
- The Experimental Ideal — CONTAINS: Random assignment, Tennessee STAR experiment, why experiments are the benchmark for causal inference
- The Selection Problem — CONTAINS: Potential outcomes framework, selection bias decomposition, overview of methods that address it
- Directed Acyclic Graphs — CONTAINS: DAG definition, forks/chains/colliders, conditioning rules, paths, backdoor paths, backdoor adjustment formula, d-separation, d-connection, valid adjustment sets, worked example with complex DAG, Simpson’s Paradox
Sources
- Mostly Harmless Econometrics.pdf — Mostly Harmless Econometrics (Angrist & Pischke, 2008), Chapters 1-2
- Unlock the Secrets of Causal Inference with a Master Class in Directed Acyclic Graphs — Graham Harrison, Towards Data Science (2023-04-06): DAGs from basics to backdoor adjustment
See Also
- Regression Foundations — Regression as the next tool after understanding selection
- Activity Bias in Advertising — Selection bias in action: observational ad measurement