Abstract:
|
Unlike traditional probability sampling, the sample size of respondent driven sampling is a random variable with little to no information about its distribution and is dependent on the number of seeds and participants' recruitment propensity. Therefore, RDS sampling productivity is neither controllable nor predictable. Low recruitment propensity will lead to the final sample size being much smaller than desired. With high recruitment propensity, one may achieve the target sample size quickly. However, if starting with a large number of seeds, recruitment chains may be too short to reach the equilibrium state. This study examines sampling productivity of RDS applied to a Web survey of foreign-born Koreans in the U.S. Specifically, we examine sample size growth over waves and recruitment chain length and whether participants' characteristics (e.g., the number of peers) and survey features (e.g., reminders) play a role. Using follow-up study data, we also examine how participants recruit their peers in practice and whether recruitment success is related to participants' recruitment strategies.
|