ROBUST CROSS VALIDATIONS IN RIDGE REGRESSION

The shrink parameter in ridge regression may be contami- nated by outlying points. We propose robust cross validation scores in ridge regression instead of classical cross validation. We use robust lo- cation estimators such as median, least trimmed squares, absolute mean for robust cross validation scores. The robust scores have global robust- ness. Simulations are performed to show the effectiveness of the proposed estimators.