Full Citation
Title: Adaptive Network Sampling (ANS): An Adaptive Network-Based Approach to Collecting Data from a Hidden Population
Citation Type: Miscellaneous
Publication Year: 2009
ISBN:
ISSN:
DOI:
NSFID:
PMCID:
PMID:
Abstract: Collecting data from a hidden population is difficult because of the absence of a sampling frame, which makes conventional survey techniques problematic. A recent variant of snowball sampling, "Respondent Driven Sampling" (RDS) has attracted considerable interest as a method of collecting data on difficult to reach populations. In this paper, we show that conventional snowball and RDS methods can be very inaccurate when sampling networks that exhibit substantial clustering due to social homophily. To solve this problem, we propose an alternative approach based on the collection of network data from respondents and an adaptive web sampling design (Thompson 2006), which combines a conventional snowball sample with a periodic "adaptive step" to jump to less explored areas of the network and guide the sample away from bottlenecks of clustered cases. In the adaptive step, we first use community detection techniques to identify clusters in the accumulated network data. Subsequently, we give sampling priority both to clusters that have been undersampled-based on an estimate of cluster size derived from the within-cluster density of sampled cases-and to cases that represent potential bridge ties to unexplored clusters. Monte Carlo results indicate that in highly clustered networks our approach is considerably more accurate than RDS and snowball sampling, resulting in a mean absolute deviation that is 80% lower than RDS samples-and close to the accuracy of random sampling-as long as the within-cluster density of ties reaches a minimum threshold.
Url: http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.473.9773&rep=rep1&type=pdf
User Submitted?: No
Authors: Mouw, Ted
Publisher: University of North Carolina, Chapel Hill
Data Collections: IPUMS USA
Topics: Population Data Science
Countries: United States