Parameter Estimation in Individual-Level Models of Infectious Disease
Spatial epidemic models are crucial to the prediction and control of infectious disease spread. Although the effect of a spatial parameter (β) on disease transmission may be more apparent in dense populations, knowledge of the spatial component of epidemic transmission is used to inform vaccination policies and culling procedures in many settings. Additionally, the susceptibility (α) of the population at risk and the infectious period (γ) affect the speed of epidemic spread. We compare the parameter estimation techniques of maximum likelihood estimation and the gold standard Bayesian MCMC in terms of width of confidence and credible intervals for all three model parameters in two epidemic frameworks. We examine the effect of misspecification of the infectious period γ on estimation of α and β. A grid population and a population generated by bivariate normal distributions are considered. We find that epidemics travel more quickly over the highly dense population regardless of the value of γ.