Respondent Driven Sampling

Before continuing on to some additional considerations for complete and partial network designs, it’s probably useful to briefly mention how respondent driven sampling is related to partial network designs. Respondent driven sampling (RDS) is an approach designed to enumerate hard to reach or hard to identify populations. RDS relies on link-tracing based recruitment that is then statistically adjusted to provide estimates of the size and characteristics of populations that are not readily enumerable otherwise (Heckathorn 1997); early examples included intravenous drug users (McKnight et al. 2006) and jazz musicians (Heckathorn and Jeffri 2001). RDS is not typically considered with other methods for gathering social network data. This is because, historically, RDS approaches rarely actually explicitly gather network data; rather RDS samples over networks. Despite relying on networks for sampling, RDS data rarely gathers and/or retains sufficient relational data to then model network characteristics of the relationships from which their sample is generated. Instead, these studies have traditionally aimed at garnering a sample of individuals who can be treated (with proper statistical adjustment) as a representative sample of a population (Gile and Handcock 2010; Goel and Salganik 2010). Recent work has proposed means for adapting RDS samples in ways that can allow for network analysis with the resulting data (Handcock and Gile 2010; Mouw and Verdery 2012; Young, Rudolph, and Havens 2018), but this has not yet become standard practice.