|
Abstract:
|
One of the important problems in epidemiology is how to measure and predict the spread of infectious diseases throughout a population of interest. There are many different ways to model the spread of diseases, with varying levels of depth and complexity. However, epidemiologists often do not consider the statistical validity of the modeling choices they make. This talk attempts to apply methods for model selection to simple infectious disease models. We focus on the epidemiological SIR compartment model, which represents the transmission of infectious diseases using ordinary differential equations to determine the number of susceptible, infected, and recovered individuals in a population at a given time. Different potential ways of estimating the parameters of the SIR model are proposed. These methods are then compared with each other alongside ordinary linear regression models, using both quantitative and visual diagnostics to determine the accuracy and precision of the estimates. We test the different methods and models on both data simulated from various known distributions and under known conditions as well as real-world data from the 2014 Ebola outbreak in West Africa.
|
