University of Glasgow
Locally Optimal Designs for Binary and Weighted Regression Models
Abstract
dc:description.abstractThe main aim of this thesis is to review and augment the theory and methods of optimal experimental designs for non-linear problems with a single variable using geometric and other arguments. It represents a continuation of the work on locally optimal designs for binary response experiments, which has been studied by Ford, Torsney, and Wu (1992) among others. Chapter 1 serves as an introduction to the non-linear design problem. The main point of difference between the non-linear case is emphasised and contrasted with the linear case. Chapter 2 presents a review of the general theory and the appropriate notation needed for the development of this thesis. Also the canonical transformation of a design problem is discussed. A necessary and sufficient condition for D-optimality of a design measure is given. Chapters 3 and 4 are devoted to the problem of constructing locally D- optimal and c-optimal designs for two parameter models respectively. In addition, the geometrical characterisation of designs optimising these criteria is discussed. Explicit solutions to compute the optimal weights of such designs are derived. Several examples of optimal designs which may be found analytically are given in chapter 3. In chapter 5 attention is focused on the problem of determining D-optimal designs for three parameter models, including those for weighted quadratic regression and generalised linear models. Chapter 6, considering the situations and problems for future work, gives a list of possible ways in which the work of this thesis may be extended.
Degree
thesis:*- Level dc:type.qualificationlevel
- MSc(R)
- Grantor dc:publisher.institution
- University of Glasgow
- Year dc:date.issued
- 1992
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Musrati, Abdelmagid K
Rights
- Language dc:language
- en