Online Program

Robust Group Variable Selection via Regularization (309988)

Keywords: Rank-based, Robust, Variable selection, Oracle

Advances in technology have led to routine collection of large-scale data in many fields such as biology, computer science, manufacturing, and so on. Variable selection, specifically group variable selection approaches, recently started to play a significant role in extracting useful information from such large-scale data. Further, data are never homogeneous and often contain outliers. In this study, we aim at selecting significant groups of variables where a model does not have within-group sparsity but has between-group sparsity and estimating the regression coefficients in a multiple linear regression model when data contain genuine outliers. Therefore, we propose a rank-based group variable selection method where a group adaptive $\ell_2$ penalty function enables us to obtain between-group sparsity. Numerical studies are conducted to assess the performance of the proposed method. Further, theoretical properties for the proposed method are also investigated.