called “M-bias” (named for the shape of the graph). Once again there is only one back-door path, and it is already blocked by a collider at B. So we don’t need to control for anything. Nevertheless, all statisticians before 1986 and many today would consider B a confounder.
No front door path?
