Directed acyclic graph


I'm a beginner at epidemiology and have an assignment where I'm trying to create a DAG using dagitty to determine what to adjust for when looking at wether country of birth has an effect on general health.

Besides the exposure (COB) and the outcome (GH) I have the following variables which I have included as variables and for which I have data in my dataset:

Sex, age, smoking, BMI, education level, economic problems.

The assignment is to figure out which confounders to adjust for but when I add all the variables, I have no arrows pointing to both COB AND GH. None of the above mentioned variables affect COB in my meaning. So my questions are

1. Isn't that the whole concept of confounding? That there is a variable that has an effect on both the outcome and the exposure?
2. When I put causal paths between the different variables I end up with a suggestion (from the software) to basically adjust for all variables in order to estimate the direct effect of COB and GH. It seems sensible but at the same time seems like I'm completely missing something.

Thanks in advance


Less is more. Stay pure. Stay poor.
So I am partially familiar with Dagitty, but have not really used it. So you are saying there are no cycles (not acyclic in your graph). Yeah, I believe if you have a confounder (d-separator?), then the software will tell you to block the backdoor?

Does everyone in your class have the same assignment, so perhaps you just aren't connecting all of the nodes that should be connected? Do you have a dataset to go with the assignment?