Answer to 5H1

[Drewniacki]

I am new to the subject, so it is more asking for explanation than reporting an issue. Hope you guys could help.

In the answer to the problem 5H1 why is DAG M -> A <- D not equivalent to original DAG M -> A -> D? I understand that conditional independence (D _||_ M | A) has to be preserved, but if M <- A -> D is considered equivalent (where M is not "connected" to D in any indirect way), why M -> A <- D is not?

Thanks for your help!