Non-Uniform Sampling in Respondent Driven Sampling (RDS)

Abstract

In this research, tweaks on RDS (Respondent Driven Sampling) are being covered. Ultimately, we are interested in reducing the variance of the prevalence. However, to this end, the estimators for the degree and contact type probability are also discussed. Two different sampling methods are being considered: uniform sampling and non-uniform sampling. For the latter, a difference can be made between perfect contact information and imperfect contact information. The use of contact information is paramount for the procedure of non-uniform sampling. And therefore, an analysis of the variance reduction of using contact information has also been investigated. This includes a simple example of a variance comparison simulation between full contact information and exclusively index case information. Included is also a description and discussion of the effects of non-uniform sampling on the attained sample, and the consequences thereof on the variance of the prevalence estimator.