Overview
Causal inference is the hardest problem in data analysis — and the most consequential. Every business decision that involves an intervention ("if we do X, Y will happen") is a causal claim. Most data analyses that support those decisions are correlational. The gap between correlation and causation is where wrong decisions live.
The right causal inference method depends on the identification strategy: what variation in the data can be used to isolate the causal effect of the treatment? Randomized experiments provide the cleanest identification. When experiments aren't possible, quasi-experimental methods — difference-in-differences, regression discontinuity, instrumental variables, propensity score matching — exploit natural variation to approximate experimental conditions. Each method has specific assumptions that must hold for the causal interpretation to be valid.
The Causal Inference Methods Prompt generates a complete causal analysis framework: identification strategy selection by data structure, assumption specification and testability, implementation with validity checks, and a results communication that correctly states the causal claim and its limitations.
What you get: - Identification strategy selection matrix - Assumption specification and testability assessment - Implementation for 4 quasi-experimental methods - Validity checks and falsification tests - Causal claim communication with correct scope
Built for: data scientists, economists, and analysts who need to make defensible causal claims from observational data — and communicate what those claims do and don't prove.