Use of the Bootstrap and Cross-Validation in Ridge Regression

摘要: Several existing methods for the choice of ridge parameter are reviewed, and a bootstrap method is proposed. The provides independent measures prediction errors based on multiple predictions along with an estimate standard error prediction. selected competitors compared through Monte Carlo simulations various degrees design matrix collinearity varying levels signal-to-noise ratio. procedure also illustrated by application to two published data sets. In one case, leads smaller mean squared than trace method. second optimal no perturbation confirmed. Benefits include its less subjective nature, ease implementation, robustness.