CMStatistics 2022: Start Registration
View Submission - CMStatistics
B1332
Title: Structure identification of space-time epidemic models Authors:  Zhiling Gu - Iowa State University (United States) [presenting]
Guannan Wang - College of William & Mary (United States)
Xinyi Li - Clemson University (United States)
Lily Wang - George Mason University (United States)
Abstract: Epidemiological models are a vital tool for understanding a course of an epidemic and making predictions. And model complexity is a crucial factor that significantly affects the forecast accuracy of pandemics. Specifically, at the early stage of a pandemic, a simple model is usually preferred due to the sparsity of cases. As the disease progresses, a more complex model with a significant amount of flexibility can better capture the heterogeneities and complexity of the underlying process. Therefore, it is of great interest to develop an analytic tool that can automatically balance the simplicity and flexibility of epidemiological models. We consider a class of space-time epidemic models (STEMs) to investigate spatial-temporal patterns in disease spread at the area level. Based on this flexible modeling framework, we develop a structure identification method to adjust the model complexity by automatically detecting predictors with linear, nonlinear, and spatially varying effects on the response. Moreover, we investigate the theoretical properties of the proposed method. We show the consistency of different types of estimators and asymptotic normality for estimators of the linear components. The proposed method is evaluated by Monte Carlo simulation studies and applied to analyze the COVID-19 outbreak.