Guan Yongtao, Yan Jun, Sinha Rajita
Division of Biostatistics, Yale University, New Haven, Connecticut 06520, USA.
Biometrics. 2011 Sep;67(3):711-8. doi: 10.1111/j.1541-0420.2011.01559.x. Epub 2011 Mar 1.
This article is concerned with variance estimation for statistics that are computed from single recurrent event processes. Such statistics are important in diagnosis for each individual recurrent event process. The proposed method only assumes a semiparametric form for the first-order structure of the processes but not for the second-order (i.e., dependence) structure. The new variance estimator is shown to be consistent for the target parameter under very mild conditions. The estimator can be used in many applications in semiparametric rate regression analysis of recurrent event data such as outlier detection, residual diagnosis, as well as robust regression. A simulation study and application to two real data examples are used to demonstrate the use of the proposed method.
本文关注的是根据单个复发事件过程计算出的统计量的方差估计。此类统计量对于每个单独的复发事件过程的诊断都很重要。所提出的方法仅对过程的一阶结构采用半参数形式，而不对二阶（即相依性）结构采用半参数形式。在非常温和的条件下，新的方差估计量被证明对于目标参数是一致的。该估计量可用于复发事件数据的半参数率回归分析中的许多应用，如异常值检测、残差诊断以及稳健回归。通过一项模拟研究和对两个实际数据实例的应用来证明所提方法的用途。
