Pretest Estimators for the Two Sample Linear Statistical Model

In this dissertation, we will be concerned with the estimation of location parameters in the linear two sample regression problem with possibly a nonscalar error covariance matrix. This problem has been examined before by many authors where identical location parameters were assumed for the two samples, and the covariance matrix had the simplist type of heteroscedasticity. Frequently however for two samples of economic data both the location and scale parameters differ. The two sample heteroscedastic model has received considerable attention recently with the development of a two stage test: one for homoscedasticity followed by a main test for the location parameters. In this context we examine the sampling characteristics of the pretest estimator that makes use of this two stage test. The analytical risk will be derived, and the UNIFORM superiority of the 2 stage pretest estimator with respect to the GAUSS MARKOV estimator is shown. Finally, an extension of the two sample heteroscedastic model is the Zellner's seemingly unrelated regression model with contemporaneously correlated errors. We will examine the pretest estimator for Zellner's seemingly unrelated regression model, and show its risk characteristics.

Name thesis:degree_name

Ph.D.

Level thesis:degree_level

Dissertation

Discipline thesis:degree_discipline

Economics

Grantor

University of Illinois at Urbana-Champaign

Year dc:date

2014

Author dc:creator

• Özçam, Ahmet

Contributors dc:contributor

• Judge, George,

• Economics, General

Identifier

(UMI)AAI8803163

OAI identifier oai:identifier

oai:www.ideals.illinois.edu:2142/70791