Lesson 3 of 6 · 12 min
Identification
State the assumptions that make a causal effect estimable from data.
An effect is identified when the assumptions let you express it with observable quantities. Common assumptions are no unmeasured confounding, positivity and consistency. Designs like difference-in-differences replace them with parallel trends.
Write assumptions next to every estimate. A reviewer should be able to say which assumption, if wrong, would change the conclusion.
Key points
- Identification is about assumptions, not estimators
- Positivity: every subgroup must have both treated and untreated units
Knowledge check
Sign in to mark this lesson complete and resume where you left off.