This simulation package is intended as a tool for experimenting with different models of social contact and social distancing as a virus spreads through a population. The virus model we use is basically an SIR model that updates daily based on simulated daily contacts between individuals. The properties of our virus model were parameterized to roughly approximate the behavior of the novel coronavirus SARS-CoV-2 as per what has been observed so far in the literature.

We used real social network data taken from a US college in 2005 to build a model of social interaction that mimicks daily life on a college campus, with weekends being more social and friend groups, housing assignment, and classyear driving the underlying probabilities of individuals interacting. The groups of close friends (there are 659) were determined using a new method of community detection we call Automated Quasi-clique Merger (AQCM) (publication in preparation, and some mathematical support currently under review for publication). The college network has 3826 individuals.

The demo video included illustrates the simulation. To begin using the package it is necessary to follow through the example simulations in the folder EXAMPLE_1. You will need to start with the file EXAMPLE_1_readme.m to understand how the modeling tools are used and how we created the example simulation. There is a "verbose version" of the main simulation script that provides more detailed documentation.

The parameters that control the behavior of the social interaction models and the virus models can be manually adjusted from inside the modeling functions. IMPORTANT TO NOTE is that THIS PACKAGE DOES NOT INTEND TO ACCURATELY PREDICT OUTCOMES IN COVID-19 SCENARIOS. The parameters in these models have been manually tuned to create a simulation that appears to be roughly correct, and good enough to roughly compare different scenarios of social distancing. We are making this package publically available at this time in order that social and medical scientists can build models for their purposes and tune them for various research purposes. In its current state the models in this package provide good examples of the differences between outcomes resulting from different models of social interaction (see EXAMPLE_1). It is for this purpose of comparing scenarios that we find the tools here most useful.