Simulate more realistic social interactions

[phelps-sg]

The model currently uses a spatial mixing model, which does not necessarily capture the scale-free degree-distribution inherent in empirical social networks. It might be relevant to explicitly model social networks, so that people are more likely to interact with a social circle and wider familiar who are not co-resident. The modeler could choose parameters of a random graph generator that were calibrated against empirical networks. Incorporating a scale-free social network could capture the effect of "super-spreaders" who could be very influential on the course of an epidemic (https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1578276/).

Socializing in bars, restaurants and night-clubs could also be explicitly simulated, as transmission could be much higher in these locations (https://wwwnc.cdc.gov/eid/article/26/7/20-0764_article). The number, size and locations of these businesses could be very relevant factors.

[robscovell-ts]

This makes sense from a heuristic/empirical point of view. In New Zealand, analysis of the most significant clusters showed that they were related to social gatherings in bars, weddings etc.