How Complete is Complete?

It is important to acknowledge that even the most complete network designs have limitations in how comprehensively they can capture the population of interest.44 These can reflect limitations as simple as less than complete response rates, question refusals, or other limitations common in any survey-based research design. It can also arise from participation refusals that are unique to network designs. For example, suppose your institutional review board requires active consent for inclusion of any information about study participants. If Steve refuses to be part of the study, but Ellen names Steve as one of her network partners, while you may have a report of that particular edge in your data, any of Steve’s characteristics, or relationships to others within the population will be unavailable to the research team. In other words, individual refusals can even limit the utility of information that is available from other sources in a network research design, making what’s available analytically less than the “complete” network aims the research began with.45