Follow-up to mDAGs—now with selection bias!

Ryan Carey, Marina Maciel Ansanelli, Elie Wolfe, and Robin Evans have released a paper characterizing the interventional equivalence class of directed acyclic graph models under both marginalization and selection (or conditioning).

They show that, unlike for the corresponding model under just marginalization (mDAGs), there are subtle differences between the ‘single-world’ intervention scheme, and weaker probing schemes (Ansanelli et al., 2024). They also provide a separation criterion that is at least as powerful as m-separation.