MSE evaluation of ridge estimators based on stochastic prior information

Abstract

Ridge regression based on stochastic prior information is examined. This formulation allows the ridge parameter k to be specified in an explicit a priori manner. After reparameterization into an exactly restricted, least squares model on an extended parameter space, ridge estimates resulting from any choice of k can be MSE evaluated using convenient tables, A numerical example is used to illustrate the developed test procedure.

Keywords:

• minimum average risk linear estimation
• mixed estimation
• multicollinearity
• ordinary least squares
• restricted least squares
• ridge regression