Statistical Methods for Outcome-Dependent Sampling

描述（由申请人提供）：我们提出了创新且具有成本效益的抽样设计，使研究人员能够以固定预算收集更多信息的样本。我们还将开发/评估新的和有效的统计方法，将收获这些设计提供的收益。将开发和传播用户友好的软件和拟议设计/方法的算法。建议的设计通常是多阶段的基础上，是一个有偏的抽样方案，其中一个观察的主要暴露变量的概率取决于结果变量和辅助协变量。拟议的研究是为了响应设计一个更强大的研究的需要，这些设计已被用于目前正在进行的研究。这些研究是PI与国家环境健康科学研究所的研究人员合作，研究环境暴露对癌症和其他疾病的影响。具体目标包括：（1）为挪威母亲和儿童队列研究（MoBA）开发FDS和两阶段FDS设计，并评估估计经验可能性方法。（2）提出了海湾石油泄漏长期跟踪研究（GuLF）的两阶段概率依赖抽样（PDS）设计，并对PDS设计的线性回归分析和线性混合模型进行了评价。将使用这些方法分析GuLF研究的数据。(3)为R代研究开发两阶段纵向结果依赖性抽样（LODS）抽样设计，采用基于基线响应的抽样方案或基于所有响应总和的抽样方案。（4）建立两阶段ODS设计和两阶段FDS设计中连续二次反应的推断程序，并对来自CPP研究、MoBa研究和铀矿研究的数据进行分析;（5）功效研究和最佳样本量分配，以实现两阶段FDS和两阶段PDS设计研究的给定预算的最大功效。每种方法的优点和缺点将通过理论研究和模拟研究进行严格审查。开发的软件将通过出版物和专用网页提供，该网页将附有“用户指南”以及如何使用这些软件的说明性数据示例。成功完成拟议的研究将对未来如何进行具有成本效益的生物医学研究以及如何有效分析这些研究的数据产生重大影响。

DESCRIPTION (provided by applicant): We propose innovative and cost-effective sampling designs that will enable the investigators to collect more informative samples at a fixed budget. We will also develop/evaluate new and efficient statistical methods that will reap the gains provided by these designs. User-friendly software and algorithms of the proposed designs/methods will be developed and disseminated. The proposed designs are generally multi-stage based and are of a biased sampling scheme where one observes the main exposure variable with a probability that depends on the outcome variable and auxiliary covariates. The proposed research is in response to the needs in design a more powerful study and these designs have been used in the current ongoing studies. These studies are collaborations the PI has with researchers at National Institute of Environmental Health Sciences to study the effects of environmental exposures on cancer and other diseases. The specific aims include: (1) Develop a FDS and two-stage FDS design for the Norwegian Mother and Child Cohort Study (MoBA) and evaluate an estimated empirical likelihood method. Data from MoBa study as well as the Cancer Risk in Uranium Miners Study will be analyzed; (2) Pro- pose a two-phase probability-dependent sampling (PDS) design for the Gulf Oil Spill Long-term Follow Up Study (GuLF) and evaluate a linear regression analysis and linear mixed model with PDS design. Data from GuLF Study will be analyzed with these methods. (3) Develop a two-stage Longitudinal outcome dependent sampling (LODS) sampling design for the Generation R Study with either a baseline response based sampling scheme or summation of all responses based sampling scheme. Data from Generation R Study as well as Collaborative Perinatal Project (CPP) will be analyzed with these methods; (4) Develop inference procedure for a continuous secondary response in a two-stage ODS design and a two-stage FDS design; Data from CPP Study, MoBa Study, and Uranium Miners Study will be analyzed; (5) Power study and optimal sample size allocation to achieve the maximum power for a given budget for studies with a two-stage FDS and two-phase PDS designs. The strengths and weaknesses of each proposed method will be critically examined via theoretical investigations and simulation studies. The developed software will be made available through publication and dedicated web page which will come with "User's Guide" as well as illustrative data examples on how to use them. Successful completion of the proposed research will have a significant impact on how future cost-effective biomedical studies to be conducted and how data from these studies be efficiently analyzed.