Skip to main content
. 2022 Jul 5;31(9):932–943. doi: 10.1002/pds.5500

FIGURE 3.

FIGURE 3

Causal diagram illustrating one scenario where the use of marginal empirical associations for confounder selection can result in over‐adjusting for instrumental variables. In this causal structure, X 2 is marginally associated with both treatment and outcome, but is independent of the outcome after conditioning on X 1.