Sensitivity Analysis of Respondent-Driven Sampling

• Activity Number: 315
• Type: Contributed
• Date/Time: Tuesday, August 6, 2013 : 8:30 AM to 10:20 AM
• Sponsor: Survey Research Methods Section
• Author(s): Sunghee Lee*+ and Tuba Suzer Gurtekin and Michael Elliott
• Companies: University of Michigan and University of Michigan and University of Michigan
• Keywords: Respondent driven sampling ; sampling rare populations

Abstract

Respondent-driven sampling (RDS) is a new sampling method specifically proposed for rare populations. Unlike traditional sampling methods which require extensive screening efforts, RDS exploits the population networks from selected seed respondents through incentivized coupons and samples from them without screening in waves. While use of RDS has increased tremendously with the pressing needs for studying rare populations, its formal base for inferences is relatively weak and sits on strong assumptions. This paper tests sensitivity of the memoryless nature of the Markov process assumed for RDS, which states that as recruitment waves continue, the sample characteristics become independent of the seeds. We use a publicly available RDS data targeting those at risk for HIV/AIDS, such as drug users and MSM through face-to-face interviews. Estimates of key variables from the full sample will be used as benchmarks and compared to those from subsamples, which will based on a subset of the seeds selected at random and nonrandom. The subsample distributions will be examined for the bias and the coverage of the nominal 95% coverage intervals as sensitivity test criteria.