Semiparametric dispersal kernels in stochastic spatiotemporal epidemic models
The dispersal kernel plays a fundamental role in stochastic spatiotemporal epidemic models. By quantifying the rate at which an infectious source infects a susceptible individual in terms of their separation distance, the dispersal kernel is able to account for the observed spatial characteristics of an epidemic. The aim of this thesis is to construct a dispersal kernel which belongs to a semiparametric family. We introduce a new concept called the natural bridge basis in order to build the semiparametrized dispersal kernel. We use data from a citrus canker epidemic in Florida to illustrate and examine our approach. We find features of the semiparametrized dispersal kernel which were not previously evident in parametrized dispersal kernels.
Advisor: Parry, Matthew
Degree Name: Master of Science
Degree Discipline: Mathematics and Statistics
Publisher: University of Otago
Keywords: Bayesian statistics; Markov chain Monte Carlo; SI model; spline method; plant disease
Research Type: Thesis