On limitations of uniplex networks for modeling multiplex contagion

Abstract: Many network contagion processes are inherently multiplex in nature, yet are often reduced to processes on uniplex networks in analytic practice. We therefore examine how data modeling choices can affect the predictions of contagion processes. We demonstrate that multiplex contagion processes are not simply the union of contagion processes over their constituent uniplex networks. We use multiplex network data from two different contexts -- (1) a behavioral network to represent their potential for infectious disease transmission using a "simple" epidemiological model, and (2) users from online social network sites to represent their potential for information spread using a threshold-based "complex" contagion process. Our results show that contagion on multiplex data is not captured accurately in models developed from the uniplex networks even when they are combined, and that the nature of the differences between the (combined) uniplex and multiplex results depends on the specific spreading process over these networks.