A talk that goes over [this paper](publication/admg-model/)
We give graphical criteria for state variables to be 'valid' in off-policy learning in a framework that generalizes dynamic treatment regimes (DTRs) and Markov decision processes (MDPs).
Causal models in statistics are often described by acyclic directed mixed graphs (ADMGs), which contain directed and bidirected edges and no directed cycles. This article surveys various interpretations of ADMGs, discusses their relations in …
Directed mixed graphs permit directed and bidirected edges between any two vertices. They were first considered in the path analysis developed by Sewall Wright and play an essential role in statistical modeling. We introduce a matrix algebra for …
We provide a unified review of various confounder selection criteria in the literature and the assumptions behind them.