Related Experiment Videos
Selecting factors predictive of heterogeneity in multivariate event time data
1Biostatistics Branch, National Institute of Environmental Health Sciences, MD A3-03, P.O. Box 12233, Research Triangle Park, North Carolina 27709, USA. dunson1@niehs.nih.gov
Biometrics
|June 8, 2004
Summary
This study introduces a Bayesian approach using gamma frailty models to test for heterogeneity in multivariate survival analysis. The methods, illustrated with lung cancer data, enable robust hypothesis testing on cluster variations.
Area of Science:
- Biostatistics
- Statistical Modeling
- Survival Analysis
Background:
- Multivariate survival analysis often requires testing for heterogeneity within clusters.
- Existing methods may not adequately address overall and class-specific heterogeneity structures.
Purpose of the Study:
- To develop a Bayesian framework for testing heterogeneity in multivariate survival data.
- To model various hypotheses about heterogeneity using a sequence of gamma frailty models.
Main Methods:
- Utilized a Bayesian approach with prior distributions for frailty variances (mixtures of point masses at zero and inverse-gamma densities).
- Employed a counting process formulation for conditional posterior distributions.
- Implemented a data augmentation Gibbs sampling algorithm for posterior computation.
Main Results:
- The prior formulation naturally allocates probability to different heterogeneity models, including homogeneity.
- A single Gibbs sampling run yields model-averaged estimates and posterior model probabilities.
- Demonstrated effectiveness using data from a lung cancer clinical trial.
Conclusions:
- The proposed Bayesian gamma frailty models provide a flexible framework for testing complex heterogeneity structures in survival data.
- The methodology facilitates robust hypothesis testing for overall and class-specific heterogeneity.
- The approach is computationally efficient and applicable to real-world clinical trial data.