Vine copula based likelihood estimation of dependence patterns in multivariate event time data
Vine copula based likelihood estimation of dependence patterns in multivariate event time data
复制标题
DOI：
10.1016/j.csda.2017.07.010
复制
发表时间：
2016-03
期刊：
影响因子：
--
通讯作者：
N. Barthel;Candida Geerdens;Matthias Killiches;P. Janssen;C. Czado
中科院分区：
文献类型：
--
作者：
N. Barthel;Candida Geerdens;Matthias Killiches;P. Janssen;C. Czado
在许多研究中，多变量事件时间数据是从相同大小的聚类中生成的。需要灵活的模型来捕捉这些数据中可能复杂的关联模式。藤系联结就是为了这个目的。对于完整的数据，可以使用葡萄系连接函数的推断方法。然而，事件时间数据经常受到右删失的影响。因此，现有的推理工具，如似然估计，需要适应。我们开发了基于似然推理的集群右删失事件时间数据使用藤蔓copula。由于右删失的单重和二重积分出现在似然表达式中，因此需要数值积分进行似然评估。一个模拟研究，灵感来自三维大鼠数据的曼特尔等人。（1977），提供了证据，所提出的方法的良好的有限样本性能。使用Laevens等人的四维乳腺炎数据。（1997），我们展示了如何为手头的数据选择一个合适的vine copula模型。我们进一步证明，我们的研究结果是符合Geerdens等人的结果。（2015），其中Joe-Hu copulas用于分析该数据集。
In many studies multivariate event time data are generated from clusters of equal size. Flexible models are needed to capture the possibly complex association pattern in such data. Vine copulas serve this purpose. Inference methods for vine copulas are available for complete data. Event time data, however, are often subject to right-censoring. As a consequence, the existing inferential tools, eg likelihood estimation, need to be adapted. We develop likelihood based inference for clustered right-censored event time data using vine copulas. Due to the right-censoring single and double integrals show up in the likelihood expression and numerical integration is needed for the likelihood evaluation. A simulation study, inspired by the three-dimensional rat data of Mantel et al.(1977), provides evidence for the good finite sample performance of the proposed method. Using the four-dimensional mastitis data of Laevens et al.(1997), we show how an appropriate vine copula model can be selected for the data at hand. We further demonstrate that our findings are in line with the results in Geerdens et al.(2015) where Joe-Hu copulas are used to analyze this data set.