Reading & Evaluation
How to Read an Empirical Paper
A systematic framework for reading causal inference papers critically. Learn to identify the estimand, evaluate identification strategies, and judge credibility.
19 Practices & Guides
Practices & Guides
Research practices, workflows, and practical skills for credible empirical work.
Research Practices
8Statistical methods and procedures organized by where they fit in your research pipeline.
1
Design Stage
2Design
Pre-Analysis Plans & Pre-Registration
Commit to your analysis before seeing the results — the antidote to the garden of forking paths.
Experimental Design
Design
Power Analysis & Sample-Size Planning
How large a sample do you need to detect your effect? Power calculations prevent underpowered studies.
Experimental DesignDifference-in-Differences (Canonical 2×2)
2
Estimation Stage
3Estimation
Randomization Inference
When conventional asymptotics fail — few clusters, unusual randomization — Fisher's exact approach provides valid inference.
Experimental DesignDifference-in-Differences (Canonical 2×2)
Estimation
Clustering and Few-Cluster Inference
When to cluster standard errors, at what level, and what to do when you have few clusters.
OLS (Robust SEs, Clustering)Difference-in-Differences (Canonical 2×2)Staggered DiDInstrumental Variables / 2SLS+3 more
Estimation
Multiple Hypothesis Testing
Testing many hypotheses inflates false positives. Bonferroni, Holm, BH-FDR, and Romano-Wolf corrections.
Experimental DesignEvent Studies (Dynamic Treatment Effects)Double/Debiased Machine Learning (DML)Causal Forests / Heterogeneous Treatment Effects
3
Robustness Stage
3Robustness
Sensitivity Analysis for Unobservables
How robust are your results to omitted variable bias? Oster (2019) and Cinelli & Hazlett (2020) provide formal answers.
OLS (Robust SEs, Clustering)Difference-in-Differences (Canonical 2×2)Fixed Effects (Two-Way FE)Matching (PSM, CEM, NN, Weighting)+23 more
Robustness
Specification Curve Analysis
How much do your results depend on the specific analytical choices you made? Explore the full space of defensible specifications.
OLS (Robust SEs, Clustering)Difference-in-Differences (Canonical 2×2)Fixed Effects (Two-Way FE)Random Effects+2 more
Robustness/Estimation
Lee Bounds for Attrition
When point identification fails — especially due to differential attrition — informative bounds can still be useful.
Experimental DesignDifference-in-Differences (Canonical 2×2)
Practical Guides
11Hands-on walkthroughs for reading papers, writing results, replicating studies, choosing methods, and avoiding common pitfalls.
Reading & Evaluation
External Validity and Generalization
When and how to generalize causal estimates beyond the study population. PATE vs SATE vs LATE, site-selection bias, reweighting for target populations.
Writing & Reporting
Writing Up Your Results
How to present causal inference results in academic papers. Covers table formatting, effect size reporting, and common writing mistakes.
Research Workflow
Natural Experiment Workflow
Decision tree for identifying and exploiting natural experiments. From finding variation to choosing estimators, running diagnostics, and reporting results.
Research Workflow
Working with Observational Data
Workflow for causal inference from observational data. Covers estimator selection, managing the assumption burden, and building credibility without experiments.
Research Workflow
Choosing Your Standard Errors
When to use heteroscedasticity-robust, clustered, two-way clustered, Conley spatial, or wild bootstrap standard errors. Decision tree with code for every SE type in R, Python, and Stata.
Research Workflow
DiD vs. Synthetic Control: When to Use Which
A practical comparison of Difference-in-Differences and Synthetic Control methods, covering assumptions, strengths, and when each is appropriate. Includes the Synthetic DiD hybrid.
Research Workflow
Matching vs. IPW vs. Doubly Robust
A practical comparison of selection-on-observables estimators: matching, inverse probability weighting, and doubly robust/AIPW methods. Covers assumptions, tradeoffs, and when to use each.
Research Workflow
When to Use DML Over Standard OLS
A practical guide to choosing between OLS regression with controls and Double/Debiased Machine Learning. Covers high-dimensional confounders, flexible functional forms, and regularization bias.
Research Workflow
How to Replicate a Study
Step-by-step guide for replicating published causal inference papers. Covers finding data, reproducing results, diagnosing discrepancies, and extending analyses.
Pitfalls & Anti-Patterns
Causal Inference Anti-Patterns
Common mistakes and anti-patterns to avoid in research design, estimation, interpretation, and reporting, with explanations and fixes.