|
Abstract:
|
Respondent-driven sampling (RDS) is a network sampling method used to access hidden populations that are not reachable by conventional sampling techniques. RDS relies on peer recruitment and thus the sampling mechanism is partially unknown to researchers and often poorly understood. Many RDS estimators rely on the assumption that recruitment occurs uniformly at random - that is, people select members of their social network at random to recruit into the study. In practice, this assumption is likely not met, and preferential recruitment occurs. It is therefore necessary to model the RDS sampling mechanism to make inferences about the population. The rational choice preferential recruitment (RCPR) model uses a two-sided utility maximization framework to model recruitment, allowing for preferential selection on observed nodal or dyadic covariates. In this talk, we assess the RCPR model using both simulated RDS data and case studies from Morocco, and discuss joint modeling of the utilities and recruit count distribution.
|
