Community-Based Studies in Cancer and Environment

描述(由申请者提供)：这是一份关于癌症和环境的社区研究相关统计问题的方法论研究的资助申请。拟议的活动是由国际癌症研究所目前参与的癌症预防和环境流行病学项目推动的。具体地说，PI计划： (1)建立空间生存数据模型。我们提出了两类模型：空间脆弱生存模型和边缘空间生存模型。本文提出了一种基于脆弱性模型的推算偏似然评分法。对于边际生存模型，我们采用复合似然方法，建立估计方程，同时估计回归系数和方差分量。 (2)研究空间生存数据的两个特殊方面。我们分析了空间失效时间数据的一般截尾类型，包括右截尾、左截尾和区间截尾。我们建立了协变量中具有测量误差的空间相关故障时间数据的模型。 (3)为在离散时间(例如，在定期、预先指定的后续时间)观察到的空间相关数据开发模型。具体地说，我们考虑用于空间相关的分组生存数据的模型和用于分析重复的离散响应的空间过渡回归模型，例如，在空间环境中观察到的二元状态。为了应对信息或不可忽视的辍学，我们还提出了联合建模分组生存和过渡过程。 经验数据分析将在所有具体目标中发挥核心作用。可用的相关数据包括工人反对烟草风险研究、东波士顿哮喘研究和老年人流行病学研究的既定人群。尽管这项研究的动机是癌症和环境健康研究中的应用问题，但该研究为生存分析、空间分析、纵向数据分析和测量误差建模的一般统计方法提供了有益的贡献。

DESCRIPTION (provided by applicant): This is a grant application for methodological research in statistical issues related to community-based studies in cancer and environment. The proposed activities are motivated by the projects of cancer prevention and environmental epidemiology that the PI is currently involved in. Specifically, the PI plans to: (1) Establish models for spatial survival data. We propose two classes of models: spatial frailty survival models and marginal spatial survival models. We develop the imputed partial likelihood score method to draw inference based on the frailty models. For the marginal survival models, we adapt the composite likelihood approach and develop an estimating equation to estimate the regression coefficients and the variance components simultaneously. (2) Study two special aspects of spatial survival data. We analyze spatial failure time data with general censoring types, including right, left and interval censoring. We develop models for spatially correlated failure time data with measurement errors in covariates. (3) Develop models for spatially correlated data that are observed in discrete time (for example, at regular, pre-specified follow-up times). Specifically, we consider models for spatially correlated grouped survival data and spatial transitional regression models for the analysis of repeated discrete responses, e.g. binary status, observed in a spatial setting. To cope with informative or non-ignorable dropout, we also propose jointly modeling the grouped survival and the transitional processes. Empirical data analysis will play a central role in all specific aims. Available relevant data include Workers Against Risk of Tobacco Study, East Boston Asthma Study and Established Populations for the Epidemiologic Study of the Elderly. Although motivated by applied problems in cancer and environmental health studies, the proposed research offers useful contributions to general statistical methodology in survival analysis, spatial analysis, longitudinal data analysis and measurement error modeling.