Bayesian variable selection for time series regression, with application to waste water monitoring
LE3 .A278 2014
Master of Science
Mathematics and Statistics
Mathematics & Statistics
Waste water treatment facilities, owned and run by municipalities and industrial entities, rely on biological agents to perform the bulk of the treatment work. The bacterial agents are sensitive to numerous factors in their environment. This thesis evaluates a speci c implementation of a Bayesian structural time series model with covariates to build predictive models for bacterial health. The model uses stochastic search variable selection (SSVS), assuming independent predictors when performing its selection. A modi cation is made to the SSVS process to induce a dependence structure on the covariates' inclusion, and its e ects are evaluated with respect to MCMC convergence and predictive power.
The author grants permission to the University Librarian at Acadia University to reproduce, loan or distribute copies of my thesis in microform, paper or electronic formats on a non-profit basis. The author retains the copyright of the thesis.