Nonparametric Estimation and Inference of a Monotone Hazard Ratio Function

Nonparametric estimation and inference of a non-decreasing monotone hazard ratio from a right censored survival dataset. The estimator is based on a generalized Grenander typed estimator, and the inference procedure relies on direct plugin estimation of a first order derivative. More details please refer to the paper "Nonparametric inference under a monotone hazard ratio order" by Y. Wu and T. Westling (2023) .

MonotoneHazardRatio

Overview

MonotoneeHazardRatio is a tool for nonparametric estimation and inference for a monotone non-decreasing hazard ratio, based on the work "Nonparametric inference under a monotone hazard ratio order" by Y. Wu and T. Westling (2022) arXiv:2205.01745.

Dependent packages

Our packages needs the following packages to work.

library(survival)
library(fdrtool)
library(KernSmooth)

Usage

It is staightforward to use this package. First you need to import the data (optional: split the data into two groups "S" and "T" such that the hazard ratio $\lambda_S/\lambda_T$ is non-decreasing). Pass the dataframes along with the evaluation grid to the function monotoneHR(), which takes $\alpha =0.05$ as the default confidence level, to have the hazard ratio and its confidence intervals estimated.

Example

As shown in the example, we are going to estimate a non-decreasing hazard ratio using the example data survData. The estimated hazard ratio is stored in theta$hr, while the confidence intervals are stored in theta$ci.lower and theta$ci.upper.

library(MonotoneHazardRatio)

### Use the example data in the package
data(survData)

### split it into two dataframes
s.data <- survData[survData$group == 'S']
t.data <- survData[survData$group == 'T']

### Evaluation grid
t.grid <- seq(0, 10, 1)

### Estimation and inference
theta <- monotoneHR(t.grid, s.data, t.data)