題 目：高維數據變量選擇的統計方法
主講人： 張敏 教授 (美國普渡大學)
時 間：2017年12月15日 上午10:30
地 點：主樓418
主講人介紹：
Dr. Min Zhang is a professor of Statistics at Purdue University. She received her MD from Hebei Medical Unviersity, PhD in Neurobiology from Peking Univeristy Health Science Center, and PhD in Biometry from Cornell University. Her current research focuses on developing statistical methods that can extract information from biomedical big data more efficiently and effectively, including methods for quantitative trait loci mapping and genome-wide association studies. Recently she is working on variable selction methods that can applied to systems biology and precision medicine.
內容介紹：
We developed a variable selection method, namely penalized orthogonal components regression, to simultaneously model multiple response variables for data with large number of predictors but small sample size. Orthogonal components are sequentially constructed to maximize their correlation to the response residuals. A new penalization framework through empirical Bayes thresholding is employed to efficiently identify sparse predictors of each component. The method can group highly correlated predictors and is computationally efficient. Extensive computer simulation studies show the superior performance of the proposed method, and it has been applied to real data collected in biomedical studies.
（承辦：管理科學與物流系，科研與學術交流中心）
