This is an R package for the treatment effect estimation using radial basis function (RBF)-nets with shared neurons based on our paper Individualized Multi-Treatment Response Curves Estimation using RBF-net with Shared Neurons (Biometrics: https://doi.org/10.1093/biomtc/ujaf019)
devtools::install_github("pshuwei/RBF")
The 'R' folder contains the coded methodology of our proposed method in Section 3 of the paper.
The 'inst' folder contains all simulation code that we discuss in the paper.
The rbf_4.1_mse.R and rbf_4.2_mse.R code contains the simulation code discussed in Section 4.1 and 4.2 regarding the simulation settings and MSE results.
The rbf_4.2_blp.R code contains the simulation code discussed in Section 4.2 regarding the second simulation and obtaining the Best Linear Projections (BLPs) for each predictor for each
The rbf_4.2_threshold.R contains the simulation code discussed in Section 4.2 regarding the thresholding inference step to identify important predictors based on the BLP.
Finally results.R is the code to run all the above code as well as obtain the MIMIC cohort ids.
Our data comes from the Medical Information Mart for Intensive Care III (MIMIC-III) data. The data are not publicly available due to privacy or ethical restrictions, but can be accessed freely by submitting an application to PhysioNet.
The 'data' folder contains the rbf_mimic_cohort.csv file which contains the subject IDs and the corresponding treatment groups we assign them based on the paper.