Date of Award
Open Access Dissertation
The cost of testing for infectious diseases can be high, and analyzing large amounts of historical data to forecast disease prevalence can be a time-consuming process. Group testing techniques offer a way to decrease the expenses associated with testing, but this comes at the cost of increased complexity of the data structure due to grouping and imperfect testing when compared to individual-level testing. In this dissertation research, we propose two Bayesian mixed effects spatial group testing regression models: one for areal data and the other for point process data. The former model utilizes a conditional autoregressive (CAR) prior for areal spatial effects, while the latter incorporates spatial information using a Gaussian predictive process. The performance of the two proposed models is evaluated through simulation studies that examine various scenarios. Subsequently, the models are applied to real data sets, namely spotted fever group Rickettsia testing data collected from ticks and West Nile virus testing data collected from mosquitoes, for the former and latter models, respectively.
Disease forecasting and surveillance often involve fitting models to a tremendous volume of historical testing data collected regularly from the specific regions of interest. Spatio-temporal mixed effects binomial regression models are commonly used for such data. In the Bayesian paradigm, these models are often fit using Markov chain Monte Carlo (MCMC) methods. However, the process of fitting these models requires significant computational resources. In this dissertation research, we propose a computationally efficient gradient boosting algorithm for fitting a Bayesian spatio-temporal mixed effects binomial regression model to perform the forecasting tasks for disease forecasting surveillance.
Huang, R.(2023). Bayesian Spatial and Spatio-Temporal Models for Complex Data Sets. (Doctoral dissertation). Retrieved from https://scholarcommons.sc.edu/etd/7515
Available for download on Saturday, August 31, 2024