Tools for understanding causal claims, directed acyclic graphs, and quasi-experimental designs — drawn from MIT, Stanford, UC Berkeley, Carnegie Mellon, and UT Austin curricula.
Paste a news headline making a causal claim. Name the treatment and outcome, identify confounders, draw a directed acyclic graph, and work through the potential outcomes framework — what would have happened to the same units without treatment?
Coordinate a causal inference reading group. Members tag the method used (DiD, RDD, IV, propensity score), sketch a DAG, vote on whether the identifying assumption holds, and build a shared library of papers worked through together.
Work through causal inference problem sets: given a real scenario (school tutoring rollout, minimum wage change, hospital admission threshold), pick the right design — RDD, DiD, IV, or synthetic controls — and articulate the identifying assumption.
Design a difference-in-differences analysis with staggered treatment timing. Articulate the parallel trends assumption in your specific case, choose the estimator family, plan heterogeneity analysis by subgroup, and produce a structured methods memo for your supervisor.
Review program evaluation reports for causal credibility: log the claimed treatment and outcome, draw the DAG, evaluate the identifying assumption, check robustness, and rate overall credibility on a structured rubric defensible to federal oversight.
