Analyzing Non-proportional Hazards: Use of the MRH Package

TL;DR

This paper demonstrates how to analyze right-censored survival data using the MRH package in R, which implements a Bayesian semi-parametric multi-resolution hazard model allowing flexible hazard estimation and covariate effects under various assumptions.

Contribution

It introduces the MRH package that enables flexible hazard rate estimation and covariate effect modeling, including non-proportional hazards, using a Polya-tree based Bayesian approach.

Findings

Effective hazard estimation in sparse failure periods

Flexible modeling of covariate effects, including non-proportional hazards

Implementation of pruning for robust hazard estimation

Abstract

In this manuscript we demonstrate the analysis of right-censored survival outcomes using the MRH package in R. The MRH package implements the multi-resolution hazard (MRH) model, which is a Polya-tree based, Bayesian semi-parametric method for flexible estimation of the hazard rate and covariate effects. The package allows for covariates to be included under the proportional and non-proportional hazards assumption, and for robust estimation of the hazard rate in periods of sparsely observed failures via a "pruning" tool.