题目: Workshop on high dimensional statistics: theory and applications
主持: Hongyuan Cao
时间: Monday, June 4, 2018
9:00-9:50am: Bing-Yi Jing, Professor of Mathematics, Hong Kong University of Science and technology
Biography: Dr. Jing is a Professor at the Department of Mathematics, Hong Kong University of Science and Technology (HKUST). He is the Director of the Center for Statistical Science, HKUST. He is an ASA fellow, an IMS fellow, an elected member of the ISI. He is a Cheung Kong Chair Professor
. He serves as Associate Editors for Canadian Journal of Statistics, Statistics and Its Interface, Journal of Business & Economic Statistics, Science in China, Journal of Data Science, Statistical Theory and Related Field. He published over 90 research papers. His research interests include Statistical Learning, Bioinformatics, Financial Econometrics, and Network Data.
Title: Community detection of sparse network
Abstract: Community detection for networks has been studied intensively in recent years. However, most methods focus on dense networks with little study on sparse networks. In this talk, we shall investigate ways to detect communities for sparse networks. Simulation results will be given to illustrate the performance of the proposed methods.
9:50-10:40am: Guangming Pan, Associate Professor of Mathematics, Nanyang Technological University, Singapore
Biography: Dr. Pan is an associate professor at Nanyang Technological University, Singapore. He obtained his PhD from University of Science and Technology of China in 2005 and joined Nanyang Technological University as assistant professor in 2008. His main research interests cover random matrix theory, high dimensional statistical inference and change point.
Title: Limiting Laws for Divergent Spiked Eigenvalues and Largest Non-spiked Eigenvalue of Sample Covariance Matrices.
Abstract: This talk is about the spiked eigenvalues of sample covariance matrices when the dimension and sample size both tend to infinity with certain rate. We show that the asymptotic distribution of the largest eigenvalue is the Tracy-Widow law and the largest non-spike eigenvalue is Gaussian distributed. We also explore its application in the factor model.
10:40-11:00am: Tea Break
11:00-11:50am: Yun Li, Associate Professor of Biostatistics and Genetics, UNC-Chapel Hill, USA
Biography: Dr. Li is an associate professor of Genetics and Biostatistics at UNC-CH. Dr. Li is a statistical geneticist with extensive experiences with method development and application on genotype imputation (developer of MaCH and MaCH-admix), genetic studies of recently admixed population, design and analysis of sequencing-based studies, analyses of multi-omics data including mRNA expression, DNA methylation and chromatin three dimensional organization. Dr. Li has been playing an active role in genetic studies of complex human traits resulting in a total of 21 GWAS and meta-analysis publications, including 16 in Science and Nature Genetics. Dr. Li has been leading multiple R01 projects on statistical method development for complex trait genetics, as well as on genetic studies of blood cell related traits in multi-ethnic cohorts. Dr. Li is also the Director for the Bioinformatics and Biostatistics Core of IDDRC (Intellectual and Developmental Disabilities Research Center). Dr. Li has received many awards and became the Thomson Reuters Highly Cited Researcher due to her high impact scientific work.
Title: Statistical Methods and Analysis of Chromatin Spatial Organization Data
Abstract: Chromosome conformation capture (3C) derived technologies have become increasingly popular to study the three dimensional (3D) structure of our genome and DNA looping regulatory interactions. However, methods to analyze data from 3C-dereived technologies are truly in their infancy. We (Xu et al 2015 PMID: 26543175) present a hidden Markov random field (HMRF) based Bayesian framework to detect long range chromatin interactions, accounting for spatial dependency, empirically non-negligible but been ignored by existing methods. We further proposed a computationally as well as statistically more efficient method (Xu et al 2016 PMID: 26969411) for the HMRF framework. We are extending the framework to model multi-tissue chromatin interactomic datasets, which improves both on false positive rate in detecting tissue specific interactions and on power to detect constitutive interactions across tissues. In addition, we present a compendium of Hi-C (genomewide version of 3C) data for 14 primary human tissues and 7 cell lines (Schmitt et al 2016 PMID: 27851967), elucidating new insights gained from DNA structure, integrated with transcriptional, regulatory, genomewide association results. Lastly, we have developed a visualization tool (Martin et al 2017 PMID: 28582503;http://yunliweb.its.unc.edu/hugin/) to facilitate efficient and effective data mining across a compendium of chromatin interaction data, and to suggest potential target genes of GWAS variants in relevant tissues.