|
Abstract:
|
In social networks, subjects are linked to one another forming structures that are usually of interest. The subjects may have a variable of interest which, in general, can only be observed after the subjects have been interviewed. Link-tracing sampling designs are commonly used to draw samples from such network populations. Besides being convenient, these sampling designs produce biased samples that make the estimation of quantities difficult. We discuss link-tracing designs in which only a fraction of relations are traced to include subjects into the study. Using simulated and real study datasets, we demonstrate how estimation for population quantities can be done.
|
