Skip to main content
. 2016 May 20;125(1):23–29. doi: 10.1289/EHP232

Figure 2.

Figure 2

Directed acyclic graph for the Granger causality model. Confounder U2 is measured and controlled, but confounder U1 is not. POLb is pollution before the outcome (O), and POLa is pollution after the outcome. If U1 is not controlled, there is a backdoor path from O to POLa, and an association would be expected. Hence, failure to find an association is evidence of a lack of confounding (i.e., no U1).