11112223333

概率统计系列学术报告:A robust fusion-extraction procedure with summary statistics in the presence of biased sources

发布人:日期:2023年03月07日 12:25浏览数:

报告题目:A robust fusion-extraction procedure with summary statistics in the presence of biased sources

报 告 人:王启华研究员(中国科学院数学与系统科学研究院)

报告时间:2023310日  14:30-15:30

报告地址:数学研究中心528

报告摘要:

Information from multiple data sources is increasingly available. However, some data sources may produce biased estimates due to biased sampling, data corruption, or model misspecification. This calls for robust data combination methods with biased sources. In this paper, a robust data fusion-extraction method is proposed. In contrast to existing methods, the proposed method can be applied to the important case where researchers have no knowledge of which data sources are unbiased. The proposed estimator is easy to compute and only employs summary statistics, and hence can be applied to many, different fields, e.g., meta-analysis, Mendelian randomization, and distributed systems. .The proposed estimator is consistent even if many data sources are biased and is asymptotically equivalent to the oracle estimator that only uses unbiased data. Asymptotic normality of the proposed estimator is also established. In contrast to the existing meta- analysis methods, the theoretical properties are guaranteed even if the number of data sources and the dimension of the parameter diverges as the sample size increases. Furthermore, the proposed method provides a consistent selection for unbiased data sources with probability approaching one. Simulation studies demonstrate the efficiency and robustness of the proposed method empirically. The proposed method is applied to a meta-analysis data set to evaluate the surgical treatment for moderate per iodontal disease and to a Mendelian randomization data set to study the risk factors of head and neck cancer.

报告人简介:

王启华,中国科学院数学与系统科学研究院研究员,博士生导师,国家杰出青年科学基金获得者,教育部长江学者奖励计划特聘教授,中科院“百人计划”入选者。曾在北京大学、香港大学任教及在深圳大学与浙江工商大学任特聘教授,先后访问加拿大、美国、德国及澳大利亚10多所世界一流大学。主要从事复杂数据经验似然统计推断、缺失数据分析、高维数据统计分析、大规模数据分析等方面的研究,出版专著三部,在The Annals of Statistics, JASA Biometrika等国际重要刊物发表论文140余篇,部分工作已产生持久的学术影响。曾主持国家自然科学基金委国家杰出青年科学基金项目、重点项目、多项面上项目,作为核心骨干成员先后参加了两项国家自然科学基金创新群体项目。是高维统计分会理事长,生存分析分会副理事长,中国现场统计研究会常务理事,中国概率统计学会常务理事,曾任或现任《中国科学》(中英文版)(2005-2012)、Electronic Research ArchiveAnn. Inst. Stat. MathBiostatistics & Epidemiology及《应用数学学报》英文版等刊物及《现代数学基础丛书》与《统计与数据科学丛书》的编委。

上一条:概率统计系列学术报告:Averaging principle for distribution dependent stochastic differential equations driven by fractional Brownian motion

下一条:概率统计系列学术报告:Asymptotic distribution theory for model averaging based on information criterion

【关闭】 打印    收藏