主 题: Bayesian Hybrid Inference and Its Application in HIV Viral Load Modeling and Computer Experiments
报告人: Gang Han (Biostatistics Texas A and M University, School of Public Health)
时 间: 2018-04-12 15:00-16:00
地 点: Room 1303, Sciences Building No. 1
Abstract（摘要）：In statistical inference, a typical method is either frequentist or Bayesian. However, with a large number of parameters and limited amount of data, both methods can have their own limitations: The frequentist method cannot take into account the information from prior distribution; the full Bayesian method may lead to biased estimates if some of the parameters do not have a prior but non-informative prior was given. In order to solve this problem, the parameters with prior knowledge and practical meanings are better treated as Bayesian, while others such as nuisance parameters and tuning parameters are better treated as frequentist. We have developed a Bayesian hybrid approach to incorporate the two types of parameters in one statistical model, and make inference about all parameters simultaneously. Compared with frequentist and full Bayesian approaches, the proposed approach can be more accurate and efficient. We illustrate this approach in an HIV study for modeling the viral load and a computer experiment analysis.
About the speaker（报告人介绍）： Dr. Gang Han is Associate Professor of Biostatistics in Department of Epidemiology and Biostatistics, at Texas A&M University, School of Public Health. He received his M.S. and Ph.D. in Statistics from The Ohio State University and B.S. in Computer Science from the Beijing University of Technology. He is a member of the American Statistical Association, American Public Health Association, and Institute for Operations Research and the Management Sciences. His research efforts have been in statistics theory and applications in Biomedical Research and Bioinformatics. His statistical research experiences lie in fields: 1, modeling time-to-event data for personalized cancer therapy, 2, frequentist and Bayesian hybrid inference for modeling HIV viral load, and 3, the design and analysis of computer experiments. His scientific research has been focusing on clinical trials, health outcome research, epidemiology, biomarker identification, comparative effectiveness research, and genome studies regarding next generation sequencing and SNP analysis, as well as statistical inferences based on U.S. national registration data. He has worked on multiple funded projects during his careers at Moffitt Cancer Center (2008-2012) and Yale University’s Department of Medical Oncology, Department of Biostatistics, and Department of Pathology (2012-2015). He is currently actively involved in external collaboration with Yale Cancer Center, Yale Medical School, and Moffitt Cancer Center, as well as internal collaboration with multiple departments in Texas A&M University.
