Skip to main content
. 2017 Apr 17;104(2):291–302. doi: 10.1093/biomet/asx009

Fig. 1.

Fig. 1.

Two directed acyclic graphs, where Inline graphic is the treatment and Inline graphic is the outcome of interest. (a) Directed acyclic graph for M-bias, where Inline graphic and Inline graphic are unobserved and Inline graphic is observed. (b) Directed acyclic graph for Z-bias, where Inline graphic is an unmeasured confounder and Inline graphic is an instrumental variable for the treatment-outcome relationship.