Sparse Additive Index Model for High-Dimensional Data with Grouped Covariates
报告人： Sijian Wang, University of Wisconsin-Madison
时间：2017-06-13 14:00 ~ 15:00
In this talk, motivated by genomic studies, we propose asparse additive-index model to integrate group information ofcovariates in the model. The method simultaneously constructs an indexfor each group and estimates the corresponding link function toconnect the index to the response. A novel constraint is proposed tosolve the identifiability issue when regularization on indexparameters is present. Our proposed method can not only identifyimportant groups, but also select important individual covariateswithin selected pathways. Furthermore, the proposed method has threegood properties: 1) It is flexible to model the nonlinear associationbetween covariates and response; 2) It automatically considers theinteractions among covariates within the same group; 3) It maydistinguish the effects of a covariate in all of groups it belongs to.We have studied the theoretical properties of the methods. The methodsare demonstrated using simulation studies and analysis on a TCGAovarian cancer dataset.
About the Speaker:
Dr. Sijian Wang is Associate Professor in the Department of Biostatistics and Medical Informatics and Department of Statistics at the University of Wisconsin-Madison. He obtained his Ph.D. in Biostatistics from the University of Michigan in 2008. His research interests includehigh-dimensional data analysis, statistical and machine learning, bioinformatics and statistical genomics, precision medicine, and survival analysis and longitudinal data analysis. He has won several paper awards by ENAR, ASA Computing Section, and ICSA, and has published over 40 papers in leading methodological and applied journals such as the Annals of Statistics, Annals of Applied Statistics, Biometrika, Biometrics, and PNAS.