Nonparametric survival analysis using Bayesian Additive Regression Trees (BART)

Rodney A. Sparapani, Brent R. Logan, Robert McCulloch, Purushottam W. Laud

Research output: Contribution to journalArticle

18 Citations (Scopus)

Abstract

Bayesian additive regression trees (BART) provide a framework for flexible nonparametric modeling of relationships of covariates to outcomes. Recently, BART models have been shown to provide excellent predictive performance, for both continuous and binary outcomes, and exceeding that of its competitors. Software is also readily available for such outcomes. In this article, we introduce modeling that extends the usefulness of BART in medical applications by addressing needs arising in survival analysis. Simulation studies of one-sample and two-sample scenarios, in comparison with long-standing traditional methods, establish face validity of the new approach. We then demonstrate the model's ability to accommodate data from complex regression models with a simulation study of a nonproportional hazards scenario with crossing survival functions and survival function estimation in a scenario where hazards are multiplicatively modified by a highly nonlinear function of the covariates. Using data from a recently published study of patients undergoing hematopoietic stem cell transplantation, we illustrate the use and some advantages of the proposed method in medical investigations.

Original languageEnglish (US)
Pages (from-to)2741-2753
Number of pages13
JournalStatistics in Medicine
Volume35
Issue number16
DOIs
StatePublished - Jul 20 2016
Externally publishedYes

Fingerprint

Regression Tree
Survival Analysis
Survival Function
Scenarios
Survival
Covariates
Hematopoietic Stem Cell Transplantation
Non-proportional Hazards
Reproducibility of Results
Simulation Study
Binary Outcomes
Transplantation
Function Estimation
Software
Stem Cells
Medical Applications
Modeling
Nonlinear Function
Hazard
Regression Model

Keywords

  • Cox proportional hazards model
  • ensemble models
  • hematologic malignancy
  • hematopoietic stem cell transplantation
  • Kaplan–Meier estimate
  • marginal dependence functions
  • nonproportional hazards
  • predictive modeling

ASJC Scopus subject areas

  • Epidemiology
  • Statistics and Probability

Cite this

Nonparametric survival analysis using Bayesian Additive Regression Trees (BART). / Sparapani, Rodney A.; Logan, Brent R.; McCulloch, Robert; Laud, Purushottam W.

In: Statistics in Medicine, Vol. 35, No. 16, 20.07.2016, p. 2741-2753.

Research output: Contribution to journalArticle

Sparapani, Rodney A. ; Logan, Brent R. ; McCulloch, Robert ; Laud, Purushottam W. / Nonparametric survival analysis using Bayesian Additive Regression Trees (BART). In: Statistics in Medicine. 2016 ; Vol. 35, No. 16. pp. 2741-2753.
@article{f61c5ff92595407d9f9dda132f450b56,
title = "Nonparametric survival analysis using Bayesian Additive Regression Trees (BART)",
abstract = "Bayesian additive regression trees (BART) provide a framework for flexible nonparametric modeling of relationships of covariates to outcomes. Recently, BART models have been shown to provide excellent predictive performance, for both continuous and binary outcomes, and exceeding that of its competitors. Software is also readily available for such outcomes. In this article, we introduce modeling that extends the usefulness of BART in medical applications by addressing needs arising in survival analysis. Simulation studies of one-sample and two-sample scenarios, in comparison with long-standing traditional methods, establish face validity of the new approach. We then demonstrate the model's ability to accommodate data from complex regression models with a simulation study of a nonproportional hazards scenario with crossing survival functions and survival function estimation in a scenario where hazards are multiplicatively modified by a highly nonlinear function of the covariates. Using data from a recently published study of patients undergoing hematopoietic stem cell transplantation, we illustrate the use and some advantages of the proposed method in medical investigations.",
keywords = "Cox proportional hazards model, ensemble models, hematologic malignancy, hematopoietic stem cell transplantation, Kaplan–Meier estimate, marginal dependence functions, nonproportional hazards, predictive modeling",
author = "Sparapani, {Rodney A.} and Logan, {Brent R.} and Robert McCulloch and Laud, {Purushottam W.}",
year = "2016",
month = "7",
day = "20",
doi = "10.1002/sim.6893",
language = "English (US)",
volume = "35",
pages = "2741--2753",
journal = "Statistics in Medicine",
issn = "0277-6715",
publisher = "John Wiley and Sons Ltd",
number = "16",

}

TY - JOUR

T1 - Nonparametric survival analysis using Bayesian Additive Regression Trees (BART)

AU - Sparapani, Rodney A.

AU - Logan, Brent R.

AU - McCulloch, Robert

AU - Laud, Purushottam W.

PY - 2016/7/20

Y1 - 2016/7/20

N2 - Bayesian additive regression trees (BART) provide a framework for flexible nonparametric modeling of relationships of covariates to outcomes. Recently, BART models have been shown to provide excellent predictive performance, for both continuous and binary outcomes, and exceeding that of its competitors. Software is also readily available for such outcomes. In this article, we introduce modeling that extends the usefulness of BART in medical applications by addressing needs arising in survival analysis. Simulation studies of one-sample and two-sample scenarios, in comparison with long-standing traditional methods, establish face validity of the new approach. We then demonstrate the model's ability to accommodate data from complex regression models with a simulation study of a nonproportional hazards scenario with crossing survival functions and survival function estimation in a scenario where hazards are multiplicatively modified by a highly nonlinear function of the covariates. Using data from a recently published study of patients undergoing hematopoietic stem cell transplantation, we illustrate the use and some advantages of the proposed method in medical investigations.

AB - Bayesian additive regression trees (BART) provide a framework for flexible nonparametric modeling of relationships of covariates to outcomes. Recently, BART models have been shown to provide excellent predictive performance, for both continuous and binary outcomes, and exceeding that of its competitors. Software is also readily available for such outcomes. In this article, we introduce modeling that extends the usefulness of BART in medical applications by addressing needs arising in survival analysis. Simulation studies of one-sample and two-sample scenarios, in comparison with long-standing traditional methods, establish face validity of the new approach. We then demonstrate the model's ability to accommodate data from complex regression models with a simulation study of a nonproportional hazards scenario with crossing survival functions and survival function estimation in a scenario where hazards are multiplicatively modified by a highly nonlinear function of the covariates. Using data from a recently published study of patients undergoing hematopoietic stem cell transplantation, we illustrate the use and some advantages of the proposed method in medical investigations.

KW - Cox proportional hazards model

KW - ensemble models

KW - hematologic malignancy

KW - hematopoietic stem cell transplantation

KW - Kaplan–Meier estimate

KW - marginal dependence functions

KW - nonproportional hazards

KW - predictive modeling

UR - http://www.scopus.com/inward/record.url?scp=84978975552&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84978975552&partnerID=8YFLogxK

U2 - 10.1002/sim.6893

DO - 10.1002/sim.6893

M3 - Article

C2 - 26854022

AN - SCOPUS:84978975552

VL - 35

SP - 2741

EP - 2753

JO - Statistics in Medicine

JF - Statistics in Medicine

SN - 0277-6715

IS - 16

ER -