Causal Inference Course

Lesson 1 • Potential Outcomes Framework

Presents the Rubin potential outcomes model as a formal basis for causal effects. Students learn to express individual and average treatment effects precisely.

Lesson 2 • Causal Assumptions and Identification

Explains what it means for a causal effect to be identified from data. Covers the core assumptions required before any estimation strategy is valid.

Lesson 3 • Directed Acyclic Graphs Basics

Introduces DAGs as a visual language for encoding causal assumptions. Students learn to read and draw simple graphs representing data-generating processes.

Lesson 4 • Correlation Versus Causation

Contrasts associational and causal claims using real examples. Establishes why standard statistical associations cannot answer causal questions without additional assumptions.

Lesson 1 • Experimental Design Principles

Covers core design choices including unit of randomization, blocking, and stratification. Students learn how design decisions affect statistical power and validity.

Lesson 2 • External Validity and Generalization

Addresses when experimental findings transport to new populations or settings. Introduces reweighting methods for improving external validity.

Lesson 3 • Threats to Experimental Validity

Identifies noncompliance, attrition, and spillover as common threats to valid inference. Students learn to detect and address each threat analytically.

Lesson 4 • Why Randomization Solves Confounding

Explains how random assignment balances observed and unobserved confounders in expectation. Connects randomization to the potential outcomes framework from Chapter 1.

Lesson 1 • Matching and Stratification

Introduces matching and stratification as nonparametric alternatives to regression adjustment. Students apply exact, nearest-neighbor, and coarsened exact matching.

Lesson 2 • Propensity Score Methods

Develops propensity score estimation and its use in weighting and matching. Students implement inverse probability weighting and diagnose overlap violations.

Lesson 3 • Confounding in Observational Data

Defines confounding precisely using DAGs and potential outcomes. Students diagnose confounding in realistic data scenarios and understand its consequences for naive estimates.

Lesson 4 • Backdoor Criterion and Adjustment Sets

Teaches the backdoor criterion as a graphical rule for valid covariate adjustment. Students identify minimal sufficient adjustment sets from a given DAG.

Lesson 1 • Two-Stage Least Squares Estimation

Teaches 2SLS as the standard IV estimator and its implementation in practice. Students compute 2SLS estimates and interpret first- and second-stage results.

Lesson 2 • IV Applications and Extensions

Surveys classic and modern IV applications including natural experiments and shift-share designs. Students evaluate instrument validity in applied research contexts.

Lesson 3 • The IV Identification Strategy

Explains how a valid instrument breaks the endogeneity problem by providing exogenous variation. Students connect IV logic to the potential outcomes framework.

Lesson 4 • Instrument Validity Testing

Covers empirical tests for instrument strength and overidentification. Students apply F-statistics, Sargan-Hansen tests, and placebo checks.

Lesson 1 • Nonparametric RD Estimation

Covers local polynomial regression and bandwidth selection as the standard RD estimation approach. Students implement and compare estimators using real data.

Lesson 2 • RD Design Logic and Setup

Explains how a cutoff in an assignment variable creates local quasi-random variation. Students identify RD-applicable settings and define the local average treatment effect at the cutoff.

Lesson 3 • Fuzzy RD and Kink Designs

Extends RD to settings with imperfect compliance and kinked assignment rules. Students implement fuzzy RD via 2SLS and interpret regression kink designs.

Lesson 4 • RD Validity Tests

Teaches density tests, covariate balance checks, and placebo cutoff tests to validate RD assumptions. Students apply each test and interpret results correctly.

Lesson 1 • Regression Implementation of DiD

Covers the two-way fixed effects regression as the standard DiD estimator. Students implement TWFE models and interpret treatment effect coefficients.

Lesson 2 • DiD Logic and the Parallel Trends Assumption

Explains how DiD removes time-invariant confounding by comparing changes across groups. Students formalize the parallel trends assumption and understand when it is plausible.

Lesson 3 • Testing and Validating DiD Assumptions

Teaches formal pre-trend tests and placebo outcome checks for DiD validity. Students apply these tests and adjust their designs based on results.

Lesson 4 • Synthetic Control Method

Introduces synthetic control as a complement to DiD for single-unit treatment cases. Students construct a synthetic control and conduct inference via permutation.

Lesson 1 • Heterogeneous Treatment Effects Overview

Introduces the concept of treatment effect heterogeneity and its policy relevance. Students distinguish average effects from conditional average treatment effects.

Lesson 2 • Estimating Mediation Effects

Covers regression-based and nonparametric methods for estimating mediation. Students implement mediation analysis and construct valid confidence intervals.

Lesson 3 • Causal Mediation Analysis

Defines direct and indirect effects using the potential outcomes framework. Students decompose total effects into natural direct and indirect components.

Lesson 4 • Policy Targeting with Heterogeneous Effects

Translates CATE estimates into optimal treatment assignment rules. Students construct and evaluate policy rules using welfare metrics.

Lesson 1 • Sensitivity Analysis for Unmeasured Confounding

Teaches Rosenbaum bounds, E-values, and partial R-squared methods to quantify robustness. Students apply these tools to assess how much hidden bias would overturn conclusions.

Lesson 2 • Interference and Network Effects

Extends causal inference to settings where units interact and SUTVA is violated. Students define exposure mappings and estimate spillover effects.

Lesson 3 • Communicating Causal Evidence

Develops skills for presenting causal findings to technical and nontechnical audiences. Students produce clear causal claims with appropriate uncertainty quantification.

Lesson 4 • Doubly Robust and Semiparametric Estimators

Introduces augmented IPW and targeted maximum likelihood as doubly robust estimators. Students implement these methods and understand their efficiency properties.