Bayesian epidemic models for spatially aggregated count data.

Area of Science:

• Epidemiology
• Biostatistics
• Statistical Modeling

Background:

• Epidemic data frequently exhibit excess zeros, spatial dependencies, and temporal correlations.
• Environmental noise and time-varying factors complicate traditional disease spread modeling.
• Accurate modeling is crucial for understanding disease dynamics and informing control strategies.

Purpose of the Study:

• To develop a flexible Bayesian regression framework for spatio-temporal count data with excess zeros.
• To incorporate serial correlation and time-varying covariates using an Ornstein-Uhlenbeck process.
• To evaluate different prior specifications and distance kernels for epidemic modeling.

Main Methods:

• Utilized a general class of stochastic regression models for spatio-temporal count data.
• Incorporated serial correlation and time-varying covariates via an Ornstein-Uhlenbeck process.
• Developed branching process-based methods for disease control scenario testing.

Main Results:

• The proposed Bayesian model effectively handles excess zeros and spatio-temporal dependencies in epidemic data.
• Exploration of various priors and distance kernels provided insights into model sensitivity.
• Branching process methods successfully linked epidemiological models with stochastic processes for policy analysis.

Conclusions:

• The developed Bayesian approach offers a robust tool for analyzing complex epidemic data.
• The methodology facilitates informed decision-making in disease control and public health policy.
• The study demonstrates the practical application of advanced statistical modeling in real-world epidemiological scenarios.