Regression Modeling Of Data Collected Using Respondentdriven Sampling
Respondent-Driven Sampling (RDS) is a snowball-type sampling method used to survey hidden populations. To date, analyses of RDS data have primarily consisted of estimating population proportions and their variance because of the special complexities RDS data pose for regression analysis. This paper discusses those complications, focusing on the role of homophily (differential affiliation) in the recruitment process and respondent clustering at multiple potential levels of aggregation. It proposes two techniques for confronting these problems: entering recruiter characteristics directly into recruit-level regression models and estimating fixed- or random-effects models at the levels where significant clustering is observed. An empirical example demonstrates the modeling process, and a six-step procedure for regression modeling of RDS data is presented.
dissertation or thesis