项目名称: 概率抽样设计及其统计推断方法
项目编号: No.11471335
项目类型: 面上项目
立项/批准年度: 2015
项目学科: 数理科学和化学
项目作者: 许王莉
作者单位: 中国人民大学
项目金额: 65万元
中文摘要: 在流行病学研究中,协变量对响应变量的影响往往是人们关心的问题之一。如果感兴趣的协变量的数据获取花费很多,在有限资金的情况下,如何获取数据使得模型的估计更有效?Zhou, Xu, Zeng和Cai(2014)提出概率抽样方法得到两阶段数据比目前常用的抽样方法更有效。 本课题致力于研究关于概率抽样方法的统计推断问题. 在感兴趣的协变量存在辅助变量的情况下,我们将考虑如何把辅助变量的信息用到概率抽样和统计推断中,充分利用数据的信息提高有效性。如果数据存在中心效应,本课题试图用概率抽样的方法在不同中心获取数据,用混合效应模型做推断。目前的概率抽样是关于线性模型的统计推断,半参数模型的概率抽样设计也是本课题研究的问题之一。现有的关于两阶段抽样数据的统计推断都是模型估计,模型检验文献中还没有涉及到,关于模型的检验也是非常重要的.
中文关键词: 概率抽样;辅助变量;随机效应模型;半参数模型;模型检验
英文摘要: In epidemiological studies, the relationship between the covariates and the response variables is an important topic. If investigating the interested covariate is too expensive, then how to get more efficient estimators by sampling under a limited budget? Zhou, Xu, Zeng and Cai (2014) proposed probability sampling designs to get two-stage data, which is more efficient than existing methods. In this project, we will mainly study statistical inference for the probability sampling method. If an auxiliary variable for the interested variable exists, we will consider how to make full use of the auxiliary variable in sampling design and statistical inference. For the case where the data set has center effect, we try to apply the probability sampling designs to each center, and use mixed effect models for inference. Probability sampling is only for linear models now, so extending to partial linear models is also one of our project themes. Existing researches on two-stage sampling data are about model estimation, and few about model checking. So it is a significant topic to do model checking based on this kind of data.
英文关键词: Probabiliy sampling designs;Auxiliary variable;Random effect model;Semi-parametric model;Model check