DAGitty

Suggested Exercise
(~10 minutes)
Taken from r-project.org.

Go to the DAG Maker page and take a moment to examine the interface. To the left, you’ll find your legend and some display options. On the right, you’ll find information related to identifying causal effects, testable implications, a code export option for R, as well as an excerpt that will generate this DAG model when pasted back into DAGitty.

Identifying Potential Confounders
Suppose you’re a researcher investigating the association between education and diabetes, and you want to know whether you should adjust for your subjects mother’s history of diabetes. This is linked to both education (through income) and your subject’s diabetes (through genetic risk). Using the tools you see here, try and create a model of these relationships.

Your graph should look something like this. Do you suspect a confounder here?

Delving Deeper
However, the association with the outcome and exposure is not direct for either: it’s due to confounding by genetic risk and childhood income, respectively. Drawing this as a DAG helps clarify that the maternal diabetes status is a collider, and that adjusting for it will induce an association between genetic risk and childhood income, which opens a back-door path from education to diabetes status. Here’s what that looks like.

What about when your study design already stratifies on m?

Helpful Literature
https://cran.r-project.org/web/packages/ggdag/vignettes/bias-structures.html