Sampling from networks: respondent-driven sampling

Abstract Objectives Respondent-Driven Sampling (RDS) is a variant of link-tracing, sampling technique for surveying hard-to-reach communities that takes advantage community members' social networks to reach potential participants. While the RDS mechanism and associated methods adjusting at analysis stage are well-documented in statistical sciences literature, methodological focus has largely been restricted estimation population means proportions, while giving little no consideration network parameters. As network-based method, faced with fundamental problem from where features such as homophily (the tendency individuals similar traits share ties) differential activity ratio average number connections by attribute) sensitive choice method. Methods Many simple approaches exist generate simulated data, specific levels (mainly activity), on estimating proportions (Gile 2011; Gile et al. 2015; Spiller 2018). However, recent findings inconsistency estimators partially observed (Crawford 2017; Shalizi Rinaldo 2013) raise question whether those target can be recovered using data alone – recovering information about these critical if we wish condition upon them. In this paper, conduct simulation study assess accuracy existing methods, terms their abilities samples desired two parameters: activity. Results The results show (1) cannot consistently estimated (2) more precise when groups, defined traits, equally active represented population. We use approach mimic Engage Study, an sample gay, bisexual other men who have sex Montréal, Canada. Conclusions highlight it possible, some cases, simulate mimicking characteristics real-world retaining precision samples.