Restricted Tree-based Reinforcement Learning

R functions and simulation study for "Estimating Tree-based Dynamic Treatment Regimes Using Observational Data with Restricted Treatment Sequences"

• TRL Functions.R stores functions for original TRL method.
• Assignx3.R stores true treatment assignment and true optimal regimes for the simulation.

/simulation_code

• simulation_true_pi_RTRL_NaivePatial.R is the main simulation code comparing the RT-RL method with naive T-RL with partial non-viable records and uses true pi models.
• simulation_wrong_pi_RTRL_NaivePatial.R is the main simulation code comparing the RT-RL method with naive T-RL with partial non-viable records and uses wrong pi models.

/additional_simulations

• simulation_true_pi_RTRL_NaiveFull.R is the main simulation code comparing the RT-RL method with naive T-RL deleting all non-viable records and uses true pi models.
• simulation_wrong_pi_RTRL_NaiveFull.R is the main simulation code comparing the RT-RL method with naive T-RL deleting all non-viable records and uses wrong pi models.