A1230
Title: A generalized estimating equation approach to network regression
Authors: Riddhi Pratim Ghosh - Bowling Green State University (United States) [presenting]
Jukka-Pekka Onnela - Harvard University (United States)
Ian Barnett - University of Pennsylvania (United States)
Abstract: Modeling the spread of infectious diseases, such as COVID-19, through a network of individuals, hospitals, or countries poses methodological challenges. Naive regression neither properly accounts for network community structure nor accounts for the dependent variable acting as both model outcome and covariate. To address this methodological gap, a network regression model is proposed, motivated by the observation that controlling for community structure can, when a network is modular, significantly account for meaningful correlation between observations induced by network connections. A generalized estimating equation (GEE) approach is proposed to learn model parameters based on node clusters defined through any single-membership community detection algorithm applied to the observed network. A necessary condition on the network size and edge formation probabilities is provided to establish the asymptotic normality of the parameters under the stochastic block framework. The approach is applied to estimate the impact of the commercial air transportation network between countries on the spread of COVID-19 incidence rates as well as on the receipt of aid between countries. Results show that while naive regression overstates the significance of network effects post-lockdown, the proposed approach more accurately quantifies the impact of the travel network on COVID-19 incidence rates.
