Brown University Center for Statistical Sciences Seminar
Abstract: We consider estimation of the mean vector of a spherically symmetric distribution with unknown scale when the mean vector is restricted to lie in a cone. A main example is the restriction of the mean vector to a polyhedral cone. This example includes such common restrictions as ordered parameters and general linear inequality constraints. We show that certain types of shrinkage estimators have the strong robustness property that they improve over the MLE for the normal model uniformly over the class of all spherically symmetric distributions.
*This work is joint with Dominique Fourfrinier and Martin Wells.
Special Scientific Computing Seminar
<--- 2003 Index