Li Yang, Sun Jianguo, Song Shuguang
Department of Statistics, University of Missouri, United States.
Support & Service Technology, The Boeing Company, United States.
Comput Stat Data Anal. 2012 Jun;56(6):2041-2050. doi: 10.1016/j.csda.2011.12.010.
This paper discusses parametric analysis of bivariate failure time data, which often occur in medical studies among others. For this, as in the case of univariate failure time data, exponential and Weibull models are probably the most commonly used ones. However, it is surprising that there seem no general estimation procedures available for fitting the bivariate Weibull model to bivariate right-censored failure time data except some methods for special situations. We present and investigate two general but simple estimation procedures, one being a graphical approach and the other being a marginal approach, for the problem. An extensive simulation study is conducted to assess the performances of the proposed approaches and shows that they work well for practical situations. An illustrative example is provided.
本文讨论双变量失效时间数据的参数分析，这类数据在医学研究及其他领域中经常出现。为此，与单变量失效时间数据的情况一样，指数模型和威布尔模型可能是最常用的模型。然而，令人惊讶的是，除了针对特殊情况的一些方法外，似乎没有适用于将双变量威布尔模型拟合到双变量右删失失效时间数据的通用估计程序。我们针对该问题提出并研究了两种通用但简单的估计程序，一种是图形方法，另一种是边际方法。进行了广泛的模拟研究以评估所提出方法的性能，结果表明它们在实际情况下效果良好。文中还给出了一个示例。
