A Simulation Study Comparing Epidemic Dynamics on Exponential Random Graph and Edge-Triangle Configuration Type...

Area of Science:

• Network Science
• Epidemiology
• Computational Social Science

Background:

• Accurate network models are crucial for understanding epidemic spread.
• Existing models like ERGMs and configuration models have different strengths in capturing network properties.

Purpose of the Study:

• To compare empirically grounded random network models (ERGMs and configuration models) for their ability to replicate network features and SIR epidemic dynamics.
• To evaluate model performance across diverse empirical contact networks.

Main Methods:

• Fitting exponential random graph models (ERGMs) and configuration model extensions to three distinct empirical contact networks.
• Developing a novel method for fitting an edge-triangle model to a snowball sampled network.
• Simulating Susceptible-Infected-Recovered (SIR) epidemic dynamics on generated networks and comparing results to empirical data.

Main Results:

• ERGMs excel at capturing network clustering, while configuration models better represent node degree distributions.
• ERGMs offer only marginal improvements in recreating epidemic features compared to configuration models, despite higher computational costs.
• Configuration model epidemic simulations generally fall between Erdős-Rényi and ERGM results.
• Adding subgraphs to edge-triangle models improved empirical network agreement for clustered networks at lower densities.

Conclusions:

• Both ERGMs and configuration models have limitations in fully capturing complex network structures and their impact on epidemic spread.
• Configuration models provide a computationally efficient alternative with comparable epidemic simulation performance to ERGMs for certain network types.
• Further research into incorporating specific network motifs, like cliques, may enhance model accuracy for networks with household structures.