Abstract
In a system of regression models, finding a feasible shrinkage is demanding since the covariance structure is unknown and cannot be ignored. On the other hand, specifying sub-space restrictions for adequate shrinkage is vital. This study proposes feasible shrinkage estimation strategies where the sub-space restriction is obtained from LASSO. Therefore, some feasible LASSO-based Stein-type estimators are introduced, and their asymptotic performance is studied. Extensive Monte Carlo simulation and a real-data experiment support the superior performance of the proposed estimators compared to the feasible generalized least-squared estimator.
Type
Publication
Statistics, Optimization and Information Computing
Add the full text or supplementary notes for the publication here using Markdown formatting.
