Seminar Series: William Whipple Neely, University of Washington

Respondent-Driven Sampling is an innovative sampling technique that has recently gained considerable popularity as a method for studying "hidden" and "hard-to-reach" populations. Furthermore, the RDS methodology comes with strategies that, it is claimed, make it possible to compute estimates of population-level characteristics and for constructing confidence intervals for such estimates. Yet despite the widespread use of RDS, there remain serious questions about the statistical validity of the methodology. This talk will explore the kinds of assumptions made in order to justify the claims of the current RDS methodology and the scientific issues that the method raises. In particular we will discuss the risks of drawing population-level inferences when these assumptions fails and whether or not there are better options for studying hidden and hard-to-reach populations.