Modelling sampling probabilities for Respondent-Driven Sampling

Respondent-Driven sampling (RDS) is a network-based sampling method devised to overcome challenges with sampling hard-to-reach human populations. Under RDS, surveyed individuals are responsible for most of the recruitment of the study participants. As such, the researchers have little control over the sampling process, the sampling probabilities are unknown, and statistical inferences highly depend on a representation of this sampling scheme. In this presentation, we discuss different approaches to model RDS that consider some sampling features (e.g., nonrandom recruitment, without replacement sampling) and network characteristics (e.g., partially directed graphs, homophily) common in RDS applications.