Semiparametric modeling of repeated measurements under outcome-dependent follow-up
Semiparametric modeling of repeated measurements under outcome-dependent follow-up
复制标题
DOI：
10.1002/sim.3496
复制
发表时间：
2009-03-15
影响因子：
2
通讯作者：
Lumley, Thomas
中科院分区：
文献类型：
--
作者：
Buzkova, Petra;Lumley, Thomas
在对受试者特定时间进行的重复测量的回归分析中，结果数据的可用性可能与过去的结果以及不在预期回归模型中的其他测量变量有关。本文提出了Lin和Ying(J.Am)的半参数回归方法的自然推广统计一下。阿索克。2001；96：103-126)，通过建立一类适应这种结果相关的随访的“反向强度比率”加权估计器。该估计具有封闭形式，根n相容，渐近正态，且不需要估计任何无限维参数。我们给出了几个模拟来证明估计器的性能，并展示了在不同程度依赖于结果相关变量的跟踪提示下的敏感性研究。我们使用了一项随机卫生服务研究的数据来说明我们的方法，该研究没有遵守预定的就诊计划。版权所有(C)2008 John Wiley&Sons，Ltd.
In regression analysis of repeated measurements that are taken at subject-specific times, the availability of the outcome data may be related to the past outcome and to other measured variables that are not in the intended regression model. In this paper we propose a natural extension of the semiparametric regression procedure of Lin and Ying (J. Am. Stat. Assoc. 2001; 96:103-126) by building a class of 'inverse-intensity-rate-ratio' weighted estimators that accommodate such outcome-dependent follow-up. The estimators have a closed form, are root n-consistent, asymptotically normal, and do not require estimation of any infinite-dimensional parameters. We give several simulations to demonstrate the estimator's performance and show a sensitivity study under follow-tip with various degrees of dependence on outcome-related variables. We illustrate our approach using data from a randomized health services research study with noncompliance to scheduled visits. Copyright (C) 2008 John Wiley & Sons, Ltd.