Nonparameteric Bounds for Evaluating the Clinical Utility of Treatment Rules - Code Repository

This repository contains the R code for the simulation studies presented in the paper: "Nonparameteric Bounds for Evaluating the Clinical Utility of Treatment Rules".

Please note: This repository contains the code to reproduce the simulation results. The code for the real-world data example (LEAP study) discussed in the paper is not included.

---

Overview

---

This project provides the code used to run the simulations comparing two strategies for calculating bounds on the effectiveness of clinical treatment rules:

1. Reduction Strategy
2. Conditioning Strategy

The primary script runs simulations across many randomly generated (but valid) probability distributions to empirically compare the performance and width of the bounds produced by these two methods.

---

File Descriptions

---

• which bounds are better.R: This is the main simulation script. It runs the core simulations described in the paper, comparing the two bounding strategies.

• helper_function.R: A utility script that is called by the main simulation. Its function is to generate valid random conditional probability distributions for a given causal graph.

• create_plots.R: This script takes the output from the simulation runs and generates figure used in the paper.

• README.txt: This file.

---

Requirements

---

• R
• The causaloptim R package.
• The parallel and pbapply, when using parallel computing.

You can install the necessary package from CRAN by running the following command in your R console:

install.packages("name_of_package")