Related Experiment Videos
A class of permutation tests for stratified survival data
1Office of Biostatistics Research, National Heart, Lung, and Blood Institute, Bethesda, Maryland 20892-7938, USA. jshih@helix.nih.gov
Biometrics
|April 21, 2001
Summary
We introduce novel permutation tests for stratified survival data, unifying advantages of existing log-rank tests. Simulations show these new methods outperform current approaches for censored and uncensored data.
Area of Science:
- Statistics
- Biostatistics
- Survival Analysis
Background:
- Stratified survival data analysis presents challenges in unifying test properties.
- Existing permutation tests may not optimally balance strata effects and sample sizes.
Purpose of the Study:
- To propose a new class of permutation tests for stratified survival data.
- To develop tests that unify the benefits of stratified and ordinary log-rank tests.
- To offer robust analysis for various censoring types.
Main Methods:
- Utilizing the Fay and Shih (1998) framework for score permutation.
- Employing shrinkage estimators for distribution functions with censored data.
- Permuting observations within strata.
Main Results:
- The proposed weighted Mann-Whitney functional test bridges stratified and ordinary log-rank tests.
- Performance is comparable to stratified log-rank with large strata effects or sample sizes.
- Performance mirrors ordinary log-rank with minimal strata effects.
Conclusions:
- The novel permutation tests offer a unified approach for stratified survival data.
- The tests demonstrate superior performance over existing methods in simulations.
- Flexibility in functional choice allows for stratified Prentice-Wilcoxon or difference-in-means tests.