dynamic treatment regime estimation for survival outcomes with dependent censoring
The simulations are designed to be run in R 3.5.2.
R packages such as survival, randomForestSRC, DTRreg, dtrSurv, tidyverse, and cowplot (for figures) should be installed before running.
install.packages("survival")
install.packages("randomForestSRC")
install.packages("DTRreg")
install.packages("dtrSurv")
install.packages("dplyr")
install.packages("tidyr")
install.packages("ggplot2")
install.packages("cowplot")
For cluster computing, run the bash script S2value.sh to submit all jobs implementing each simulation setting on Slurm.
To run the very first setting of the simulations (or a specific setting with slight manipulation of the code),
run C21.simulation_run.R.
source("C21.simulation_run.R")
Once all simulations are implemented, update the lab.date in C22.summary.R and run the whole script.
The leukemia data are not publicly available, and the code (L1, L2, LF) is provided for transparency.
The ARIC data are available through https://sites.cscc.unc.edu/aric/distribution-agreements. The code R1 is for data preprocessing (R1.R, R1_2.R, R1_3.R in the order), R2 is for application of the method and the competitors, and R3 is for graphical illustration of the results obtained in R2.