Regression Methods for Respondent-Driven Sampling Data

• Activity Number: 14 - Novel Statistical Methods for Network-Based Studies Among People Who Use Drugs
• Type: Invited
• Date/Time: Sunday, July 28, 2019 : 2:00 PM to 3:50 PM
• Sponsor: National Institute on Drug Abuse-NIH
• Author(s): Miles Ott*
• Companies: Smith College
• Keywords: respondent-driven sampling; networks; social network data

Abstract

In the respondent-driven sampling (RDS) survey design, data is collected through several distinct sampling trees. The use of RDS allows for the sampling of hard-to-reach populations, such as people who inject drugs, as long as those populations are densely connected networks. While RDS allows for an expedient means of attaining a sample, there are methodological issues that arise from employing RDS. In particular: independence can no longer be assumed, and unequal sampling probabilities must be accounted for. In this paper we propose an extension of generalized estimating equations method that may be applied to multivariable regression using data collected through RDS.