Improving Inferences from RDS Data by Incorporating Geo-Reference Information

• Activity Number: 251
• Type: Contributed
• Date/Time: Monday, August 10, 2015 : 2:00 PM to 3:50 PM
• Sponsor: Survey Research Methods Section
• Author(s): Breda Munoz* and Mario Chen and Kate MacQueen
• Companies: RTI International and FHI 360 and FHI 360
• Keywords: RDS ; spatial statistic model ; survey based estimators

Abstract

Respondent-driven sampling (RDS) is often used to investigate populations for which a probabilistic sample cannot be efficiently generated. RDS has become a common approach for recruiting hard to reach populations. Because RDS is not a probability sample design, standard design-based estimation approaches are not applicable. To obtain survey based estimators, researchers make a number of assumptions about the recruitment process and the underlying social network in order to calculate inclusion probabilities. We borrowed from the spatial statistics literature to improve the prediction of prevalence estimates at the regional level. Our proposed approach can be used in the identification of spatial patterns of the population of interest. This is possible by attaching participant geo-reference info to the data (e.g. latitude-longitude of the recruiting center, mailing or surrogate address of participants). This approach can be useful to evaluate the spatial distribution of the sample; to detect and characterize clusters of cases. We illustrate this approach with data collected by FHI360 in a study of black population at risk for HIV in Durham, NC.