Nonparametric methods for factorial designs with censored data

• 成果类型: Article
• 署名作者: Akritas, MG; Brunner, E
• 署名单位: University of Gottingen
• 刊物名称: JOURNAL OF THE AMERICAN STATISTICAL ASSOCIATION
• ISSN/ISSBN: 0162-1459
• DOI: 10.2307/2965705
• 发表日期: 1997
• 页码: 568-576
• 关键词: kaplan-meier statistics; regression-models; survival analysis; large sample; estimator; tests; functionals

摘要

Nonparametric hypotheses of no main effects, no simple effects, and no interaction effects are considered in the context of factorial designs with censored observations. To test these hypotheses, the asymptotic distribution of quadratic forms of rank statistics is derived using the methodology of martingales for counting processes. The weights used reduce to the usual logistic scores with uncensored data but are different from the weights commonly used for testing the equality of k samples. The formulation of all results includes tied observations. Approximations to the small-sample distributions of the test statistics are given. The performance of the tests is examined via simulation studies. The procedures are illustrated on a real dataset.