Respondent-driven sampling (RDS) is widely used to obtain estimates of quantities such as HIV prevalence, but is also used to examine correlates of infection, and less frequently, to help characterise social networks of individuals in “hidden” populations. The popularity of RDS stems in part from the often rapid recruitment of individuals from the target population as well as from the potential to obtain (asymptotically) unbiased estimates using information only from the sample. I will review the assumptions required for this potential to be realised, whether these assumptions are broken in practice, and the impact of breaking assumptions on statistical inference. I will also discuss improvements to the design and analysis of RDS studies, including schemes for giving out coupons, obtaining absolute sizes of at-risk populations using capture-recapture, and the utility of detailed social network inventories.
Statistics from Altmetric.com
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.