Measuring Homophily in Network Data

• Activity Number: 668
• Type: Topic Contributed
• Date/Time: Thursday, August 8, 2013 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Sergiy Nesterko*+ and Joseph Blitzstein
• Companies: Harvard University and Harvard University
• Keywords: homophily ; RDS ; social networks ; network sampling

Abstract

In network applications ranging from online social networks to friendship networks of hard-to-reach populations, researchers cannot observe the network directly and must instead use sampling to estimate full-network quantities of interest. We study how well different homophily indices, including the new Relative Homophily Index, estimate the true value in such settings by considering common network sampling procedures via simulation. We find that Respondent-Driven Sampling is practical and effective for recovering homophily, a result which we also justify theoretically for a class of link-tracing mechanisms.