Sparse Groupwise Envelope Model for Response Variable Selection in Imaging Genetic Analysis (306436)*Yeonhee Park, Medical University of South Carolina
Zhihua Su, University of Florida
Hongtu Zhu, University of North Carolina at Chapel Hill
Keywords: Dimension reduction, Envelope model, Grassmann manifold, Variable selection.
This paper aims to develop a sparse groupwise envelope model for response variable selection and efficient estimation under multivariate linear regression. When the observations belong to several groups, the sparse groupwise envelope model can accommodate distinct regression coefficients and heteroscedastic error structures for different groups. Consistency and the oracle property of the proposed method are established. Simulation studies and the analysis of two data sets from the Alzheimer's Disease study and the Philadelphia Neurodevelopmental Cohort show the effectiveness of our method in efficient estimation and response variable selection. Data for the Alzheimer's Disease study used in preparation of this article were obtained from the Alzheimer's Disease Neuroimaging Initiative database.