Date of Award

1988

Document Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

Department

Economics

First Advisor

R. Carter Hill

Abstract

The dissertation addresses three issues in the use of Stein-like estimators of the classical normal linear regression model. The St. Louis equation is used to generate out-of-sample forecasts using least squares. These forecasts are compared to those produced by restricted least squares, pretest, and members of a general family of minimax shrinkage estimators using the root-mean-square error criterion. Bootstrap confidence intervals and ellipsoids are constructed which are centered at least squares and James-Stein estimators and their coverage probability and size is explored in a Monte Carlo experiment. A Stein-like estimator of the probit regression model is suggested and its quadratic risk properties are explored in a Monte Carlo experiment.

Pages

353

Share

COinS