Awesome

Paper Repository

This repository contains the code to reproduce the numerical results in our paper: Conformalized Survival Analysis.

Overview

We implement the CQR- and CDR-LPB from scratch to reproduce the numerical results in Section 4 of the paper. We do not include the code for the results in Section 5 because the UK Biobank data is only available to registered users; see https://www.ukbiobank.ac.uk/ for details.

Based on the code, we further develop an R package cfsurvival that implements the procedure. The package will be constantly improved and udpated. We recommend the users download the R package to apply our procedure.

The following R packages are required to be installed: conTree, GauPro, gbm, quantreg, survival, tidyverse.

Folders

• R/: contains the main functions that implement the CQR, CDR, and other competing methods considered in Section 4.
• utils/: contains the helpers for the experiments.
• simulation/: contains the scripts to carry out the simulations.
• bash/: bash files to run the simulations in batch mode.

Usage

Single run

Each script in the simulation/ folder implements one run of the simulation. The users can specify the random seed when running the script. For example, to implement one run of the low-dimensional-homoscedastic-noise experiment in Section 4 with random seed 1, run the following command in your terminal:

mkdir results
cd simulation
Rscript ld_homosc.R 1

Multiple runs

The results presented in the paper are based on 100 independent samples. The user can use the bash file in bash/ to automatically run the whole simulation in batch mode:

mkdir results
cd bash
bash run_all.sh

It may take a long time if it is run on a laptop.