Related Experiment Videos
Tests of independence for bivariate survival data
1Office of Biostatistics Research, National Heart, Lung, and Blood Institute, Bethesda, Maryland 20892, USA.
Biometrics
|December 1, 1996
Summary
We developed two new statistical tests for bivariate survival data independence. The supremum test is powerful, while the weighted test offers flexibility for detecting specific survival data dependencies.
Area of Science:
- Biostatistics
- Survival Analysis
- Statistical Independence
Background:
- Assessing independence in bivariate survival data is crucial for accurate analysis.
- Existing methods may lack power or flexibility in detecting various dependence structures.
Purpose of the Study:
- To introduce novel test statistics for bivariate survival data independence.
- To evaluate the asymptotic properties and power of these new tests.
Main Methods:
- Proposing two test statistics derived from the covariance process of martingale residuals.
- Deriving asymptotic properties under the null hypothesis of independence.
- Investigating the asymptotic distribution and optimal weights for the weighted test.
Main Results:
- The supremum test demonstrates strong power, comparable to existing methods, especially under Clayton's family alternatives.
- The weighted test shows increased power when optimal weights are selected for specific alternatives.
- Simulations confirm the performance of the proposed tests against the Clayton and Cuzick (1985) Savage scores test.
Conclusions:
- The proposed supremum and weighted tests offer valuable tools for assessing independence in bivariate survival data.
- These tests provide alternatives with potentially improved power and flexibility depending on the underlying dependence structure.