7月5日 - 6日：明德法学楼0302
7月7日 - 8日：逸夫第二报告厅
Professor Peter Song (University of Michigan)
Dr. Peter Song is the Professor of Biostatistics at the UM School of Public Health. His research interests have primarily focused on copula models, composite likelihood, estimating equations, high-dimensional data, longitudinal data analysis. He also experts in biomedical research, such as biomarkers, chronic diseases, disease surveillance, injury, kidney paired donation program and so on. Dr. Song has become a member of editorial board of many famous journals, including Journal of the American Statistical Association, Canadian Journal of Statistics, Statistica Sinica, Journal of Planning and Inference, and several published papers have had important influence in both methodology area and biomedical area.
Title: Fusion Learning and Confederate Inference
Abstract: This talk is intended to offer a systematic introduction to data analytics and statistical algorithms in data integration. As data sets of related studies become more easily accessible, combining data sets of similar studies is undertaken in practice to achieve Big Data and to enjoy more powerful analysis. A major challenge arising from integrated data analytics pertains to principles of information aggregation, learning data heterogeneity, algorithms for model fusion. Information aggregation has been studied extensively by many statistics pioneers, which lay down the foundation of data integration. Also, ignoring such heterogeneity in data analysis may result in biased estimation and misleading inference. Four topics will be covered in the talk, including：Introduction, Seven Perspectives of Meta Estimation, Method Of Divide-And-Combine (MODAC) and Confederate Inference, Distributed and Integrated Method of Moments (DIMM).
Professor Jianhui Zhou (University of Virginia)
Dr. Jianhui Zhou is the Associate Professor at the University of Virginia. He received BS in Mathematics in University of Science and Technology of China, and Ph.D in University of Illinois at Urbana-Champaign. His areas of research have primarily focus on dimension reduction, robust statistics, longitudinal data analysis, functional linear models, semiparametric generalized linear models and so on. Jianhui Zhou has published paper in several journals, including Journal of the American Statistical Association, Annals of Statistics, Statistica Sinica, Journal of Multivariate Analysis, etc.
Title: Applying Functional Data Analysis to Children Growth Curve Modeling
Abstract: Functional data analysis is often used in health related studies when repeated measurements are intensively observed on subjects, or when a covariate can be considered as a function on some domain. Children’s growth curves of height or weigh are functions of age by nature, and we apply the functional data analysis to characterize different patterns of growth from a cohort of Bangladesh children. In this application, we develop an index to quantify the severity of growth faltering of a child compared to the WHO standard, using functional principal component analysis, and the risk factors of growth faltering can be identified using the developed index. In practice, growth measurements, such as height, are usually sparsely and irregularly taken on the underlying growth curves due to subject late enrollment or early dropout. Estimation of the growth curves over the entire study period is needed to identify the hidden growth patterns or to conduct subsequent analysis such as clustering and classification. In another application, we develop a procedure of estimating growth curves from sparsely and irregularly observed height data for a cohort of preschool children from age 2 to 18. For interpretability, our estimated curves of height should be increasing and leveling off in the end of the study period. To achieve those interpretable features of growth curves of height, we adopt a monotone transformation and a penalized estimator to ensure the level-off trend. Application of functional data analysis to other health studies, such as the study of diarrhea effects on children growth, will also be presented in this talk.
Professor Jianhua Hu (Columbia University)
Dr. Jianhua Hu is the Professor of Biostatistics at Columbia University. Her research focus on methodology development to address unconventional data analysis challenges in biomedical studies to improve disease diagnosis, prognosis, and treatment. This includes analyzing high-dimensional genomics/proteomics, imaging, and longitudinal data, modeling disease heterogeneity, and developing innovative adaptive designs to achieve personalized treatments. Dr. Hu's research substantially involves variable selection, classification, dimension reduction, nonparametric estimation, and robust inference.
Title: Statistical development for high-throughput bioinformatics data
Abstract: Various bioinformatics data have been encountered in biomedical research. Developing appropriate and effective statistical analytical tools is essential in extracting useful information from such data to make meaningful scientific discovery. Unfortunately many classical methods, such as analyzing one genetic marker at a time, are inapplicable in many applications. I will describe several problems, and the motivation and new statistical development for each of them.
Professor Hongjian Zhu (University of Texas)
Dr. Hongjian Zhu is the assistant professor at Coordinating Center for Clinical Trials, The University of Texas School of Public Health. He received BS in Statistics from Zhejiang University and Ph.D in Statistics from University of Virginia. His areas of research have primarily focused on clinical trials, adaptive designs, sequential monitoring, large sample theory, stochastic process, etc. Hongjian Zhu has published papers in several journals, including Journal of the American Statistical Association, Canadian Journal of Statistics, The Annals of Statistics and so on.
Title: Fundamentals of the interim analysis of clinical trials
Abstract: Clinical trials are the gold standard for evaluating the effectiveness of interventions. The interim analysis is one of the reliable, rational approaches to clinical trials that incorporate what is learned during a clinical study without compromising the validity or integrity. Conducting interim analysis in clinical trials has ethical, administrative and economic advantages. This talk is comprised of three parts. First, an overview of clinical trials, which begins with essential steps of clinical trials and critical sections of clinical trial protocol, will be introduced. Second, fundamentals of interim analysis including different stopping boundaries, spending functions, stochastic curtailment, inference following sequential monitoring, sample size re-estimation, and others, will be discussed. Third, recent advances in the sequential monitoring of adaptive randomized clinical trials will be offered.
Professor Judy Wang (George Washington University)
Dr. Judy Wang is the Associate Professor at George Washington University. She received BS and MS in Statistics from Fudan University, and Ph.D in Statistics from University of Illinois at Urbana-Champaign. Her areas of research have primarily focused on quantile regression, extreme value theory and applications, bioinformatics, nonparametric (semiparametric) regression, inference, variable selection, survival analysis, longitudinal data analysis, measurement error, missing data analysis. Dr. Wang has published papers in several journals, including Annals of Statistics, Journal of the American Statistical Association, Biometrika, etc.
Title: Censored Quantile Regression for Survival Analysis
Abstract: Censored quantile regression offers a valuable supplement to Cox proportional hazards model for survival analysis. It relaxes the proportionality constraint on the hazard and allows for modeling heterogeneity of the data. Moreover, the quantiles of the survival time are directly interpretable. In this talk, I will review various inference methods for censored quantile regression under different censoring schemes including both fixed and random censoring, and illustrate the implementation in R through examples.
Doctor Frank Liu (Merck Sharp & Dohme., Inc.)
Dr. Frank Liu is distinguished scientist at Merck Sharp & Dohme, Inc (MSD). He received BS in Mathematics and MS in Statistics from East China Normal University, and Ph.D in Statistics from UCLA. After completing a post-doc at Biostatistics in the Johns Hopkins University, he has been working at MSD for more than 22 years. Within MSD, he has been leading working groups to develop many technical guidance documents for statistical methods in clinical trials. He is Fellow of American Statistical Association (ASA), and member of ASA, ICSA, and Biometrics society; and has been actively publishing papers and book chapters, presenting and teaching short courses in professional meetings.
Title: Issues and Challenge for Missing Data Handling in Clinical Trials
Abstract: Missing data are inevitable and post many issues and challenge in analysis for clinical trials. Despite a great amount of research has been devoted to this topic, properly handling missing data in clinical trials remains complex. Conventionally, under the missing at random (MAR) assumption, we often use maximum likelihood or multiple imputation based methods for inferences. However, the MAR assumption is unverifiable and has recently been considered as overly-simplistic and unrealistic by regulatory agencies. In this talk, we will first review the ICH E9 addendum which advocates the use of estimand framework, and suggests several strategies to handle missing data in clinical trials. After providing an overview of conventional missing data handling methods, we will discuss several recently-developed methods, such as sensitivity analysis to assess robustness, control-based imputation, trimmed mean method, and tipping point analysis. Real clinical trial examples will be presented for illustration with implementation of the analysis using SAS software.
Professor Tingting Zhang (University of Virginia)
Dr. Tingting Zhang is the Associate Professor at University of Virginia. She received BS in Mathematics from Beijing University, MS and Ph.D in Statistics from Harvard University. Her research lies mainly in the multidisciplinary field of human brain mapping. She has collaborated closely with scientists outside statistics to develop new statistical models and computational algorithms for human brain research. Dr. Zhang has published papers in several journals, including Journal of Royal Statistical Society Series B, Journal of the American Statistical Association, NeuroImage, etc.
Title: Spatial Temporal Analysis of Multi-subject Neuroimaging Data
Abstract: Functional magnetic resonance imaging (fMRI) is one of the most popular neuroimaging technologies used for studying the human brain's activity because it provides non-invasive measurements of the entire brain's activity with a high spatial resolution. However, due to the massive size, considerable noise, and complex spatial and temporal features of fMRI data, the ensuing data analysis faces several challenges, including extensive computation and difficulty in obtaining statistically efficient estimates of the brain responses. We propose a new statistical model and computational algorithm to address these challenges. Specifically, we develop a new multi-subject, low-rank model within the general linear model framework for stimulus-evoked fMRI data. The new model assumes that the brain responses of different brain regions and subjects fall into a low-rank structure and can be represented by a few principal functional shapes. As such, the new model enables borrowing information across subjects and regions and increasing the ensuing estimation efficiency of brain responses, while accommodating the variation of brain activities across subjects, stimulus types, and regions. We propose two different optimization functions and a new fast-to-compute algorithm to address two research questions of broad interest in psychology studies: evaluating brain responses to different stimuli and identifying brain regions with different responses. Through both simulation and real data analysis, we show that the new method can outperform the existing methods by providing estimates of brain responses with much smaller variances.
Professor Jiong Li (Aarhus University, Denmark)
Dr. Jiong Li is an associate professor at Aarhus University in Denmark. He received BS at Shanghai Medical University, MSc at Erasmus University Rotterdam (The Netherlands), PhD at Aarhus University (Denmark). He has work experience in China, the Netherlands, Australia, and Denmark. Dr. Li is recognized as the No. 1 researcher in population bereavement research and one of the influential epidemiologists in register-based research. His research findings were reported multiple times by BBC, New York times, Reuters, etc. He is the grantee of the most prestigious research grant (European Research Council grant) and the main supervisor for PhD students from Denmark, Spain, England, and China. He has published more than 100 original scientific papers, including some as the first author in New Engl J Med, Lancet, Circulation, Plos Med, etc.
Title: A Gold Mine for Health Research-The Unique National Registers in Nordic Countries
Abstract: When an entire country is a cohort----- The Danish government has accumulated on its citizenry, which today totals about 5 million people. Denmark has earned a preeminent reputation for possessing the most complete and interwoven collection of statistics touching on almost every aspect of life. The Danish government has compiled nearly 200 databases, some begun in the 1930s, on everything from medical records to socioeconomic data on jobs and salaries. What makes the databases a plum research tool is the fact that they can all be linked by a 10-digit personal identification number, called the CPR, that follows each Dane from cradle to grave. The registers allow for instant, large cohort studies that are impossible in most countries, except other Scandinavian countries that have created powerful database systems. There are more than 80 medical databases maintained by the Danish Board of Health and public hospitals. Other 120 demographic databases are overseen by the agency Statistics Denmark. Working the health databases can yield powerful results. The health databases have proven invaluable for probing contradictions raised by smaller studies and following disease progression. The health databases are also useful for unraveling complex diseases. The ability to track related individuals in the many different databases makes it possible to shed light on the complex interplay between familial predisposition and environment. As an example, I am going to use studies from myself and some colleagues in Nordic countries to show, what types of data are available for health research, how the data is used, the strengths and limitations, the potential challenges, and most important, the promising perspectives to dig this gold mine.
北京市海淀区中关村大街59号 100872 中国人民大学统计学院
陈虹 010–6251 4065 firstname.lastname@example.org