Numerical construction of K-optimal designs for linear, nonlinear, and generalized linear models

dc.contributor.authorZhang, Xiaoqing
dc.contributor.supervisorZhou, Julie
dc.date.accessioned2025-12-09T21:24:53Z
dc.date.available2025-12-09T21:24:53Z
dc.date.issued2025
dc.degree.departmentDepartment of Mathematics and Statistics
dc.degree.levelMaster of Science MSc
dc.description.abstractThis thesis investigates the numerical construction of K-optimal designs for a variety of statistical models. These include linear models such as polynomial, trigonometric, and second-order response models, nonlinear models such as Michaelis–Menten, compartmental, and Peleg models, and generalized linear models with a particular focus on logistic regression. K-optimality aims to minimize the condition number of the Fisher information matrix to improve the numerical stability in parameter estimation. A general algorithm is proposed and applied to all models to construct K-optimal designs, evaluated under different design spaces and parameter values. For nonlinear models, the K-optimal designs are compared with A-optimal and D-optimal designs, while for the logistic regression model, comparisons are made with D-optimal designs. The results show that K-optimal designs have stable patterns between different models. Factors such as design space, model type, and parameter values influence the support points, their weights, and the condition number. In addition, K-optimal designs achieve smaller condition numbers, indicating better numerical stability, and take less computation time than both D-optimal and A-optimal designs. All key findings are presented in tables and figures, and the MATLAB code used for the computations is provided in the thesis.
dc.description.scholarlevelGraduate
dc.identifier.urihttps://hdl.handle.net/1828/22964
dc.languageEnglisheng
dc.language.isoen
dc.rightsAvailable to the World Wide Web
dc.subjectOptimal design
dc.subjectK-optimal design
dc.titleNumerical construction of K-optimal designs for linear, nonlinear, and generalized linear models
dc.typeThesis

Files

Original bundle
Now showing 1 - 1 of 1
Loading...
Thumbnail Image
Name:
Zhang_Xiaoqing_MASc_2025.pdf
Size:
1.65 MB
Format:
Adobe Portable Document Format
License bundle
Now showing 1 - 1 of 1
No Thumbnail Available
Name:
license.txt
Size:
1.62 KB
Format:
Item-specific license agreed upon to submission
Description: