Published in Andrew Heiss's blog

In my post on generating inverse probability weights for both binary and continuous treatments, I mentioned that I’d eventually need to figure out how to deal with more complex data structures and causal models where treatments, outcomes, and confounders vary over time.

References