Bias-Variance and Breadth-Depth Tradeoff in Respondent-Driven Sampling

• Activity Number: 629
• Type: Contributed
• Date/Time: Thursday, August 4, 2011 : 8:30 AM to 10:20 AM
• Sponsor: Section on Survey Research Methods
• Author(s): Sergiy O. Nesterko*+ and Joseph Blitzstein
• Companies: Harvard University and Harvard University
• Keywords: RDS ; Volz-Heckathorn ; hard-to-reach population ; visualization ; inference

Abstract

Respondent-driven sampling (RDS) is a link-tracing network sampling strategy to collect data from hard-to-reach populations (for example, injection drug users or individuals at high risk of being infected with HIV). Currently, the so-called Volz-Heckathorn (VH) estimator is used for RDS in practice, which uses very strong assumptions about underlying social network and RDS process. Via simulation, we study relative performance of the plain mean and VH estimator when assumptions of the latter are not satisfied, under different network types (including homophily and rich-gets-richer networks), participant referral patterns and varying number of coupons. The main question of interest is, given that the VH estimator is derived under strong simplifying assumptions, how strong a performance loss does it incur when RDS scenarios are more realistic? We convey the intuition using novel interactive visualizations of RDS process and simulation findings, created using the programming language for dynamic visualization called Processing.