Simon D.W. Frost: Respondent-driven sampling as an epidemic

Respondent-driven sampling (RDS) involves asking individuals to recruit peers into a study; in this way, the number of participants can increase potentially exponentially, with relatively little effort from the researchers conducting the study. RDS has become popular not only because of the potentially rapid recruitmen, but also that under certain assumptions, asymptotically unbiased point estimates of population composition can be obtained. I consider the recruitment process as analogous to the spread of an epidemic, and will present analyses based on multitype branching processes and stochastic epidemiological models to challenge some of these strong assumptions, as well as to extract useful information regarding the underlying social network over which RDS operates. Finally, I present analyses of an RDS study of a 'known' population, which highlights further needs for methodological research.

[Joint work with Art Poon, Yakir Berchenko, Jonathan Rosenblatt, Cyprian Wejnert, and Richard White]