项目名称: 生物医学数据统计分析的方法、理论与应用
项目编号: No.11331011
项目类型: 重点项目
立项/批准年度: 2014
项目学科: 数理科学和化学
项目作者: 王启华
作者单位: 中国科学院数学与系统科学研究院
项目金额: 240万元
中文摘要: 本项目研究生物医学数据包含高维或超高维数据、函数型数据、纵向数据、缺失数据及随机删失数据统计分析的方法、理论与应用,开展统计学与生物医学交叉研究。具体地,我们探讨生物医学一些重要问题中因果路径上多变量之间的因果作用的可识别性与传递性以及主分层的因果作用的统计推断方法;研究复杂生物数据的模型选择和模型平均方法,建立最优模型平均估计的理论;研究复杂生物试验最优设计与数据质量监控技术;发展变量选择方法、模型选择方法及降维技术分析处理生物、医学研究中高维数据及复杂高维数据;发展模型探测技术,研究函数多项式模型,使用奇异分解方法研究函数型数据的分析方法,使用相对误差准则发展函数乘积模型的统计推断方法与模型选择方法;基于一些半参数模型研究随机删失下纵向数据统计分析的方法。
中文关键词: 高维数据;函数型数据;纵向数据;缺失数据;随机删失数据
英文摘要: This project is to develop analyzing methodes and theories for biological and medical data such as high dimensional data, functional data, longitudianl data, missing data and random censoring data, and carry out the interdisciplinary research between statistics and bioscience and medical science. Specifically, we shall investigate the topics on identifiability and transitivity of causal effects among variables on causal paths, and further we shall discuss these of causal effects based on principal stratification. The related statistical inference will be considered. We intend to investigate model selection and model averaging approaches for complex biological data, and establish the optimal model average estimation theory. We shall study optimal designs for complex biological experiments and the related techniques for monitoring data quality. Variable selection, model selection approaches and dimension reduction techniques will be developed to analyze high dimensional data and complex high dimensional data in biology and medical study. We shall investigate functional polynomial regression models by developing model detection techniques. Functional singular component analysis technology will be used to investigate analyzing methods for functional data. Statistical inference approaches and model selection methods will be developed for functional product models. Some statistical methods for longitudinal data under random censorship will be investigated for some semiparametric models.
英文关键词: High dimensional data ;Functional data;Longitudinal data;Missing data;Random censoring data