Tom Britton, 
Mathematical Statistics,
Stockholm University


Title: 
Respondent driven sampling: a survey



Abstract: The traditional way of estimating a population mean or
fraction is to take a random sample in the community and to take the
corresponding sample mean/fraction as the estimate. Some populations are
however not possible to take a random sample from, so-called "hidden" or
"sensitive" populations (e.g. intraveneous drug-users). An alternative
way is then Respondent Driven Sampling, in which a few seeds are
selected. They report their response anonymously and also pass on the
questionaire to some of their "friends" in the population of interest.
The sample is hence obtained by randomly walking around in the
population network. In the talk we present this method and how estimates
can be obtained. We also illustrate why the method does not work when
the network is directed (as is often the case).