Multiplex networks involve multiple relational types across the same set of nodes, accounting for the different types of ties we have with others. Some of these ties overlap with one another, while others are more likely to be complementary in nature. Generally, the approach described in Chapter 2 recommended specificity in the way we gather tie data.69
By following such guidelines, we would have separate edge sets (or layers) corresponding to each relationship. To illustrate, see Figure 3.7, which includes two layers corresponding to different relationship types.70 For example, the “Project 90” data described above includes four relationship types (social, sexual, drug sharing, needle sharing). In examples like this, study aims may lead us to aggregate across the multiplex relationships. To analyze potential STI transmission through the Project 90 population, both sexual and needle sharing relationships are jointly informative (Rothenberg et al. 1995) and could be combined (adams, Moody, and Morris 2013). However, if we begin with only the collapsed set, we cannot subsequently dis-aggregate them from one another—e.g., if we wanted to model the potential impact of introducing a needle exchange program into that population. In that scenario, we would not be able to distinguish ties for which the intervention would reduce transmissibility (needle ties) from those that would not be altered (sexual relationships).