Comparison of hazard rate estimation in R

Yolanda Hagar, Vanja Dukic

Published 2015-09-10Version 1

We give an overview of eight different software packages and functions available in R for semi- or non-parametric estimation of the hazard rate for right-censored survival data. Of particular interest is the accuracy of the estimation of the hazard rate in the presence of covariates, as well as the user-friendliness of the packages. In addition, we investigate the ability to incorporate covariates under both the proportional and the non-proportional hazards assumptions. We contrast the robustness, variability and precision of the functions through simulations, and then further compare differences between the functions by analyzing the "cancer" and "TRACE" survival data sets available in R, including covariates under the proportional and non-proportional hazards settings.