Shrinkage methods enhanced the accuracy of parameter estimation using Cox models with small number of events

I. Feng Lin, Wushou Peter Chang, Yi Nan Liao

Research output: Contribution to journalArticle

16 Citations (Scopus)

Abstract

Objective: When the number of events is small during Cox regression analysis, it is unclear what alternative analytical strategies can be used and when this type of alternative approach is needed. This study explores several analytical strategies in this situation. Study Design and Setting: Simulations and sensitivity analyses were performed on data with numbers of events per predictive variable (EPVs) below 10 using a Cox model with a partial likelihood (PL), Firth's penalized likelihood, or the Bayesian approach. Results: For scenarios involving binary predictors with an EPV of six or less, the simulations showed that the Firth and Bayesian approaches were more accurate than was PL. The performances of various approaches were similar when the EPV was greater than six in the binary predictor. Furthermore, the performances involving continuous predictors were similar, regardless of the EPV. The bias and precision of the parameter estimates using Bayesian analysis depended on the selection of priors. Conclusions: When the EPV is six or less, the results for categorical predictors tend to be too conservative. Firth's estimator may be a good alternative in this situation. Appropriate choices of priors when using Bayesian analysis should increase the accuracy of the parameter estimates, although this requires expertise.

Original languageEnglish
Pages (from-to)743-751
Number of pages9
JournalJournal of Clinical Epidemiology
Volume66
Issue number7
DOIs
Publication statusPublished - Jul 2013

Fingerprint

Bayes Theorem
Proportional Hazards Models
Regression Analysis

Keywords

  • Bayesian methods
  • Cancer risk assessment
  • Cox proportional hazard model
  • Firth's penalized likelihood
  • Partial likelihood
  • Small number of events

ASJC Scopus subject areas

  • Epidemiology

Cite this

Shrinkage methods enhanced the accuracy of parameter estimation using Cox models with small number of events. / Lin, I. Feng; Chang, Wushou Peter; Liao, Yi Nan.

In: Journal of Clinical Epidemiology, Vol. 66, No. 7, 07.2013, p. 743-751.

Research output: Contribution to journalArticle

Lin, I. Feng ; Chang, Wushou Peter ; Liao, Yi Nan. / Shrinkage methods enhanced the accuracy of parameter estimation using Cox models with small number of events. In: Journal of Clinical Epidemiology. 2013 ; Vol. 66, No. 7. pp. 743-751.
@article{7b3e64562c7a4a8484ac4b2fc1a8c953,
title = "Shrinkage methods enhanced the accuracy of parameter estimation using Cox models with small number of events",
abstract = "Objective: When the number of events is small during Cox regression analysis, it is unclear what alternative analytical strategies can be used and when this type of alternative approach is needed. This study explores several analytical strategies in this situation. Study Design and Setting: Simulations and sensitivity analyses were performed on data with numbers of events per predictive variable (EPVs) below 10 using a Cox model with a partial likelihood (PL), Firth's penalized likelihood, or the Bayesian approach. Results: For scenarios involving binary predictors with an EPV of six or less, the simulations showed that the Firth and Bayesian approaches were more accurate than was PL. The performances of various approaches were similar when the EPV was greater than six in the binary predictor. Furthermore, the performances involving continuous predictors were similar, regardless of the EPV. The bias and precision of the parameter estimates using Bayesian analysis depended on the selection of priors. Conclusions: When the EPV is six or less, the results for categorical predictors tend to be too conservative. Firth's estimator may be a good alternative in this situation. Appropriate choices of priors when using Bayesian analysis should increase the accuracy of the parameter estimates, although this requires expertise.",
keywords = "Bayesian methods, Cancer risk assessment, Cox proportional hazard model, Firth's penalized likelihood, Partial likelihood, Small number of events",
author = "Lin, {I. Feng} and Chang, {Wushou Peter} and Liao, {Yi Nan}",
year = "2013",
month = "7",
doi = "10.1016/j.jclinepi.2013.02.002",
language = "English",
volume = "66",
pages = "743--751",
journal = "Journal of Clinical Epidemiology",
issn = "0895-4356",
publisher = "Elsevier USA",
number = "7",

}

TY - JOUR

T1 - Shrinkage methods enhanced the accuracy of parameter estimation using Cox models with small number of events

AU - Lin, I. Feng

AU - Chang, Wushou Peter

AU - Liao, Yi Nan

PY - 2013/7

Y1 - 2013/7

N2 - Objective: When the number of events is small during Cox regression analysis, it is unclear what alternative analytical strategies can be used and when this type of alternative approach is needed. This study explores several analytical strategies in this situation. Study Design and Setting: Simulations and sensitivity analyses were performed on data with numbers of events per predictive variable (EPVs) below 10 using a Cox model with a partial likelihood (PL), Firth's penalized likelihood, or the Bayesian approach. Results: For scenarios involving binary predictors with an EPV of six or less, the simulations showed that the Firth and Bayesian approaches were more accurate than was PL. The performances of various approaches were similar when the EPV was greater than six in the binary predictor. Furthermore, the performances involving continuous predictors were similar, regardless of the EPV. The bias and precision of the parameter estimates using Bayesian analysis depended on the selection of priors. Conclusions: When the EPV is six or less, the results for categorical predictors tend to be too conservative. Firth's estimator may be a good alternative in this situation. Appropriate choices of priors when using Bayesian analysis should increase the accuracy of the parameter estimates, although this requires expertise.

AB - Objective: When the number of events is small during Cox regression analysis, it is unclear what alternative analytical strategies can be used and when this type of alternative approach is needed. This study explores several analytical strategies in this situation. Study Design and Setting: Simulations and sensitivity analyses were performed on data with numbers of events per predictive variable (EPVs) below 10 using a Cox model with a partial likelihood (PL), Firth's penalized likelihood, or the Bayesian approach. Results: For scenarios involving binary predictors with an EPV of six or less, the simulations showed that the Firth and Bayesian approaches were more accurate than was PL. The performances of various approaches were similar when the EPV was greater than six in the binary predictor. Furthermore, the performances involving continuous predictors were similar, regardless of the EPV. The bias and precision of the parameter estimates using Bayesian analysis depended on the selection of priors. Conclusions: When the EPV is six or less, the results for categorical predictors tend to be too conservative. Firth's estimator may be a good alternative in this situation. Appropriate choices of priors when using Bayesian analysis should increase the accuracy of the parameter estimates, although this requires expertise.

KW - Bayesian methods

KW - Cancer risk assessment

KW - Cox proportional hazard model

KW - Firth's penalized likelihood

KW - Partial likelihood

KW - Small number of events

UR - http://www.scopus.com/inward/record.url?scp=84878248081&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84878248081&partnerID=8YFLogxK

U2 - 10.1016/j.jclinepi.2013.02.002

DO - 10.1016/j.jclinepi.2013.02.002

M3 - Article

VL - 66

SP - 743

EP - 751

JO - Journal of Clinical Epidemiology

JF - Journal of Clinical Epidemiology

SN - 0895-4356

IS - 7

ER -