Approximate minimax estimation in linear regression: a simulation study

Abstract

Let the parameter vector of the ordinary linear re-gression model be constrained to a given ellipsoid. Incorporating this kind of prior information into the estimation procedure, we obtain a sequence of approxi-mate minimax estimators. It can be shown that a cluster point of this sequence is an exact minimax estimator. In a simulation study this minimax estimator is com-pared with the least squares estimator and some other competing estimators.

Keywords:

• regression analysis
• linear model
• prior information
• minimax estimation
• nonlinear constraints