Email Record: Proper Inference for Value Function in High-Dimensional Q-Learning for Dynamic Treatment Regimes