021-55664888

扫描二维码关注
复旦大学管理学院

查看更多

统计学系系列讲座之269-270期

 

统计学系系列讲座之269

 

时间:2017 年 7月12日(星期三)10:30-11:30

地点:史带楼204室

主持人:夏寅 青年研究员 复旦大学管理学院统计学系

主题:CHIME: Clustering of High-Dimensional Gaussian Mixtures with EM Algorithm and Its Optimality

主讲人:Professor Tony Cai(蔡天文)

The Wharton School, University of Pennsylvania

个人简介:蔡天文教授是沃顿商学院Dorothy Silberberg 统计学讲席教授,中组部千人计划B入选者。2008年荣获世界统计学考普斯奖 (COPSS Presidents' Award), 该奖为国际统计学领域的最高奖。2006年当选为国际数理统计学会会士(Fellow)。他曾担任The Annals of Statistics的主编,也曾担任Journal of the American Statistical Association,Annals of Statistics,Statistica Sinica等国际著名统计期刊的编委(Associate Editor),现任Journal of the Royal Statistical Society, Series B的编委(Associate Editor)。其主要研究方向为:高维统计推断 (High dimensional statistical inference)、大范围假设检验 (Large-scale multiple comparisons)、非参数函数估计 (Nonparametric function estimation)等。

摘要:Unsupervised learning is an important problem in statistics and machine learning with a wide range of applications. In this paper, we consider clustering of high-dimensional Gaussian mixtures and pro- pose a procedure, called CHIME, that is based on the EM algorithm and a direct estimation method for the sparse discriminant vector. Both theoretical and numerical properties of CHIME are studied. We first obtain the rates of convergence for the estimation and mis- clustering error and then provide matching minimax lower bounds. The results together establish the optimality of CHIME as well as the proposed estimator of the discriminant vector. Numerical studies show that CHIME outperforms the existing methods under a variety of settings. 

 

统计学系系列讲座之270

 

时间:2017 年 7月14日(星期五)16:00-17:00

地点:复旦大学管理学院思源教授楼726室

主持人:朱仲义 教授 复旦大学管理学院统计学系

主题:Quantile Based eQTL Discovery

主讲人:Ying Wei, Ph.D 

Department of Biostatistics, Mailman School of Public Health, Columbia University 

个人简介:Dr. Ying Wei's research interests are in the general area of quantile regression, longitudinal data, and semi-parametric models, with a focus on developing methodologies of longitudinal growth chart construction. This screening process can provide an individual's current growth status by taking into account one's personal profiles. Dr. Wei's methodologies provide flexibility by avoiding underlying distribution assumption and accommodate unequally-spaced measurement time spacings. She also has investigated effective methods to make inferences, diagnose model goodness-of-fit, and assess uncertainty of screening based on the estimated models.

摘要:Over the past decade, there has been a remarkable improvement in our understanding of the role of genetic variation in complex human diseases, especially via genome-wide association studies(GWAS). However, the underlying molecular mechanisms are still poorly characterized, impending the development of therapeutic interventions. Identifying genetic variants that influence the expression level of a gene, i.e. expression quantitative trait loci (eQTLs), can help us understand how genetic variants influence traits at the molecular level. That is crucial to understand the GWAS results. While most eQTL studies focus on identifying mean effects on gene expression using linear regression, evidence suggests that genetic variation can impact the entire distribution of the expression level. Motivated by identifying the potential higher order associations, we propose to use quantile rank-score test combining over multiple quantile levels to identify eQTLs that are associated with the conditional quantile functions of gene expression. The proposed test has well controlled type I errors, and is computationally simple. We have applied this approach to the Genotype-Tissue Expression project, an international tissue bank for studying the relationship between genetic variation and gene expression in human tissues, and discovered a set of new eQTLs with heterogeneous effects across different quantile levels. We found that those heterogeneous eQTLs identified by quantile regression are associated with greater enrichment in genome-wide significant SNPs from the GWAS catalog, and are also more likely to be tissue specific than eQTLs identified by linear regression. The results suggest that quantile regression has great potential for new discoveries in genetics. 

 

统计学系

2017-7-10

快速链接

扫一扫,欢迎关注
复旦管院官方微信