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 analysis of binary prediction tree models for retrospectively sampled outcomes.

Jennifer Pittman1, Erich Huang, Joseph Nevins

  • 1Institute of Statistics & Decision Sciences, Duke University, Durham, NC 27708-0251, USA. jennifer@stat.duke.edu

Biostatistics (Oxford, England)
|October 12, 2004
PubMed
Summary

This study introduces Bayesian classification trees for case-control studies with many predictors, like gene expression data. The novel approach enhances prediction accuracy by incorporating retrospective designs and using Dirichlet process priors.

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

It Mite Not Be Graft-Versus-Host Disease: Frequency and Clinical Features of Demodex Folliculitis After Allogeneic Hematopoietic Cell Transplantation.

Transplantation and cellular therapy·2026
Same author

Improving representativeness in trial recruitment: A data-driven approach.

Contemporary clinical trials communications·2026
Same author

Author Correction: Phenome-wide analysis of copy number variants in 470,727 UK Biobank genomes.

Nature·2026
Same author

A Phase 1 Dose-Escalation Trial of Reirradiation for Patients With Recurrent Glioblastoma.

Practical radiation oncology·2026
Same author

Phenome-wide analysis of copy number variants in 470,727 UK Biobank genomes.

Nature·2026
Same author

Rare heterozygous missense variants in VSX2 are associated with retinal detachment.

PLoS genetics·2026

Area of Science:

  • Statistics
  • Bioinformatics
  • Computational Biology

Background:

  • Classification tree models offer flexibility in analyzing predictor interactions and generating response predictions.
  • Retrospective case-control designs present unique challenges, especially with a high number of candidate predictors.
  • Gene expression data studies frequently encounter scenarios with numerous potential predictors.

Purpose of the Study:

  • To describe a Bayesian analysis framework for classification tree models tailored to retrospective case-control designs.
  • To address the complexities of analyzing high-dimensional predictor data, such as gene expression data.
  • To develop and evaluate novel tree models that explicitly incorporate retrospective study designs and utilize nonparametric Bayesian methods.

Main Methods:

Related Experiment Videos

  • Bayesian analysis of classification trees with binary response data from retrospective case-control studies.
  • Application of nonparametric Bayesian models, including Dirichlet process priors, for predictor variable distributions.
  • Forward generation of trees using Bayes' factor based tests for significant binary partitions.
  • Bayesian model averaging for combining multiple trees to improve prediction.

Main Results:

  • Introduction of tree models that explicitly handle retrospective designs.
  • Demonstration of the utility of Dirichlet process priors in high-dimensional settings.
  • Successful application in predicting breast tumor status using high-dimensional gene expression data, showcasing exploratory and predictive capabilities.

Conclusions:

  • The proposed Bayesian classification tree approach effectively handles retrospective case-control data with many predictors.
  • The methodology provides a robust framework for prediction and exploratory analysis in fields like genomics.
  • Further research can extend the modeling and computational aspects of this approach.