Rubin-Style Methods
Instrumental Variables
An instrument can isolate treatment variation that is plausibly unrelated to unmeasured confounding.
Page 1 of 7 · Big Picture
The big picture: Instrumental Variables
Can outside encouragement identify an effect when treatment is confounded?
Why a policy analyst should care
An instrument can isolate treatment variation less affected by confounding, but only under strong assumptions and often for a particular subgroup influenced by the instrument.
Follow the idea from the opening question to a practical decision. These are the conceptual moves this module will develop:
1 Question What identifies an effect when treatment is confounded?
2 Explore Relevance and exclusion are separate claims
3 Formalize The Wald ratio scales outcome change by treatment change
4 Transfer Defend an instrument in health policy
Use the left and right arrow keys to move between pages.