1
$\begingroup$

Suppose we have Treatment T and Outcome O, and variables A and B, with the following causal dag:[

But if we were doing an RCT with these treatment and outcome variables, is it accurate to think of the causal dag then turning into: enter image description here

Where the incoming arrows to T are gone?

Improve this question
$\endgroup$
1
  • $\begingroup$ Do downvoters have any specific reason for doing so? Has this question been asked? $\endgroup$
    – da7666
    Commented Sep 29 at 17:20

1 Answer 1

Reset to default
2
$\begingroup$

Yes, that is correct. A and B no longer affect T, because T is determined by randomisation that is independent of A and B. Some people might put another node on the graph for 'randomisation', either to make the point or to allow the graph to represent imperfect compliance/adherence

Improve this answer
$\endgroup$
2
  • $\begingroup$ Thanks for the response. When you say " Some people might put another node on the graph for 'randomisation',", is this node have an outgoing arrow going towards the treatment? or its just like a root nodes with no children? $\endgroup$
    – da7666
    Commented Sep 30 at 23:49
  • $\begingroup$ Randomisation ->Treatmeant $\endgroup$
    – Thomas Lumley
    Commented Sep 30 at 23:52

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.