Inequality constrainted models for the multivariate normal mean: A Bayesian approach
Joris Mulder
Department of Methodology and Statistics
Faculty of Social Sciences
Utrecht University
P.O. Box 80140, 3508 TC Utrecht
Phone: +31 30 253 1227 / 4438 (secretary)
E-mail: Joris Mulder
Project: Project at Utrecht University
Project running from: 1 September 2006 – 1 September 2010
Supervisors: Prof. dr H. Hoijtink (Utrecht University), dr ir G.J.A. Fox (Twente University), dr I.G. Klugkist (Utrecht University)
Summary:
Instead of focussing on null hypothesis testing, a researcher is often interested in the alternative situation in which the null hypothesis does not hold. The researcher may have several competing theories that have to be tested with the collected data. Competing theories on the means can be, for instance, that the means are increasing from time point 1 to 4 or that the means at time point 1 and 2 are smaller than the means at time point 3 and 4. The aim of this research project is to address this issue for models for the multivariate normal mean using a Bayesian approach. Furthermore, issues that a researcher encounters such as missing data, adjustment of a constrained model, and model averaging are addressed.
Date of defence: 3 December 2010
Title of thesis: Bayesian model selection for constrained multivariate normal linear models