Jove
Visualize
Contact Us
JoVE
x logofacebook logolinkedin logoyoutube logo
ABOUT JoVE
OverviewLeadershipBlogJoVE Help Center
AUTHORS
Publishing ProcessEditorial BoardScope & PoliciesPeer ReviewFAQSubmit
LIBRARIANS
TestimonialsSubscriptionsAccessResourcesLibrary Advisory BoardFAQ
RESEARCH
JoVE JournalMethods CollectionsJoVE Encyclopedia of ExperimentsArchive
EDUCATION
JoVE CoreJoVE BusinessJoVE Science EducationJoVE Lab ManualFaculty Resource CenterFaculty Site
Terms & Conditions of Use
Privacy Policy
Policies

Related Experiment Videos

Bayesian information criterion for censored survival models.

C T Volinsky1, A E Raftery

  • 1AT&T Labs, Florham Park, New Jersey 07932, USA. volinsky@research.att.com

Biometrics
|April 28, 2000
PubMed
Summary

We revised the Bayesian Information Criterion (BIC) for censored survival data by adjusting its penalty term. This modification improves variable selection accuracy and predictive performance in survival models.

Related Concept Videos

You might also read

Related Articles

Articles linked to this work by shared authors, journal, and citation graph.

Sort by
Same author

BIC extensions for order-constrained model selection.

Sociological methods & research·2022
Same author

Consistency for the tree bootstrap in respondent-driven sampling.

Biometrika·2020
Same author

Predicting relapse prior to transplantation in chronic myeloid leukemia by integrating expert knowledge and expression data.

Bioinformatics (Oxford, England)·2012
Same author

Bayesian melding for estimating uncertainty in national HIV prevalence estimates.

Sexually transmitted infections·2008
Same author

Progress and challenges in modelling country-level HIV/AIDS epidemics: the UNAIDS Estimation and Projection Package 2007.

Sexually transmitted infections·2008
Same author

Assessment of response bias in mild head injury: beyond malingering tests.

Journal of clinical and experimental neuropsychology·2002

Area of Science:

  • Statistics
  • Biostatistics
  • Survival Analysis

Background:

  • The Bayesian Information Criterion (BIC) is widely used for model selection.
  • Standard BIC approximates Bayes factors with unit-information priors.
  • BIC's application to censored survival data requires adaptation.

Purpose of the Study:

  • To revise the Bayesian Information Criterion (BIC) for improved variable selection in censored survival data models.
  • To enhance the approximation of Bayes factors for survival data.
  • To improve the predictive performance of survival models.

Main Methods:

  • Proposed a revised BIC penalty term based on uncensored events, not total observations.
  • Applied the revised BIC to the Cox proportional hazards regression model using maximized partial likelihood.

Related Experiment Videos

  • Evaluated the revised BIC using the Cardiovascular Health Study dataset for stroke risk prediction.

    • Main Results:

    • The revised BIC penalty term provides a better approximation to the exact Bayes factor for censored data.
    • Defining BIC using the number of deaths in the Cox model penalty term aligns with a more realistic prior.
    • The revised BIC demonstrated improved predictive performance in stroke risk assessment.

    Conclusions:

    • The proposed revision of the BIC penalty term enhances its utility for variable selection in censored survival data.
    • This adaptation leads to more accurate model selection and better predictive accuracy.
    • The revised BIC offers a valuable tool for analyzing survival data, particularly in clinical and epidemiological studies.