Lee Seung-Hwan, Yang Song
Department of Mathematics and Computer Science, Illinois Wesleyan University, Bloomington, IL 61702, USA.
Lifetime Data Anal. 2007 Sep;13(3):371-80. doi: 10.1007/s10985-007-9039-z. Epub 2007 May 26.
In this paper, new statistical tests for the censored two-sample accelerated life model are discussed. From the estimating functions using integrated cumulative hazard difference, stochastic processes are constructed. They can be described by martingale residuals, and, given the data, conditional distributions can be approximated by zero mean Gaussian processes. The new methods, based on these processes, provide asymptotically consistent tests against a general departure from the model. A graphical method is also discussed. In various numerical studies, the new tests performed better than the existing method, especially when the hazard curves cross. The proposed procedures are illustrated with two real data sets.
本文讨论了删失两样本加速寿命模型的新统计检验方法。利用积分累积风险差的估计函数构建了随机过程。这些随机过程可用鞅残差来描述，并且在给定数据的情况下，条件分布可用零均值高斯过程来近似。基于这些过程的新方法针对模型的一般偏离提供了渐近一致的检验。还讨论了一种图形方法。在各种数值研究中，新检验方法比现有方法表现更好，尤其是在风险曲线交叉的情况下。所提出的程序用两个真实数据集进行了说明。
