In this talk, I will introduce the optimal dynamic treatment regimes. A dynamic regime is a function that takes treatment and covariate history to the current time as arguments and returns the treatment to be given. I will outline the steps to finding an optimal regime, with reference to the methods proposed by Murphy (2003) and Robins (2004). I demonstrate connections between the methods and difficulties that can arise in each. Interesting features will be highlighted through simulation and in the context of the Multicenter AIDS Cohort Study.
Groupe d’études et de recherche en analyse des décisions