ADAPTS-HCT-AIME

This is the repository for the paper Reinforcement Learning on AYA Dyads to Enhance Medication Adherence.

The structure of the repository is as follows:

1. Code/ contains the code for the different candidate algorithms and for the dyadic environment (in subdirectories Algorithms/ and Env/ respectively).
2. Experiment_Test_Algs/ contains the experiments to obtain and plot the cumulative rewards under different possible algorithms.
3. Experiment_Tune_Ctreat/ contains the experiments to tune the hyperparameter $C_\text{Treat}$ which controls the treatment effects and is imputed to obtain the STEs of 0.15, 0.3, and 0.5.
4. Experiment_Test_Opt_Policy/ contains the experiments to test the different optimal policy approximation candidates.
5. Model_Fitting/ contains the coefficients fitted through GEE for the dyadic environment models.
6. Opt_Policy/ contains the pickle files for the optimal policy approximation run under different environments.

Each of the directories Code/, Experiment_Test_Algs/, Experiment_Tune_Ctreat/, and Experiment_Test_Opt_Policy/ contains further detail on the code structure and running instructions.

ROADMAP Dataset was used to fit the simulator models in the project. The dataset is available for download here.