Dr. Glymour has kindly provided the attached DAG as a way to display emulation failure 4. She notes that this DAG represents the bias as resulting from the common cause of past events on exposure and future rates (and therefore more aligns with what we call confounding rather than collider bias/selection bias). However, we think this is the (or a) right DAG for the situation. She may cover this further in her upcoming lecture and feel free to ask more questions about this then.
