NEW SURVEY QUESTIONS AND ESTIMATORS FOR NETWORK CLUSTERING WITH RESPONDENT-DRIVEN SAMPLING DATA.

Area of Science:

• Social Network Analysis
• Epidemiology
• Computer Science

Background:

• Respondent-driven sampling (RDS) is widely used for sampling hard-to-survey populations, primarily for public health prevalence estimates.
• Traditional RDS methods do not typically collect data suitable for analyzing social network structures.
• Network clustering is a key topological property influencing diffusion processes like disease spread.

Purpose of the Study:

• To introduce novel data collection instruments and estimators for analyzing network clustering using RDS.
• To adapt and evaluate computer science-based network clustering estimators for RDS samples in realistic field settings.
• To expand the utility of RDS beyond prevalence estimation to include social network structural analysis.

Main Methods:

• Development of data collection instruments and RDS estimators for network clustering.
• Simulation studies to assess estimator performance under various RDS sampling conditions (e.g., multiple seeds, recruitment biases, imperfect response rates).
• Comparison of estimator behavior in RDS samples versus traditional random walk samples.

Main Results:

• Clustering coefficient estimators demonstrate robustness and retain desirable statistical properties when applied to RDS samples.
• The study identifies how specific RDS sampling characteristics impact the accuracy of network clustering estimates.
• The proposed methods successfully enable network characteristic calculations from nontraditional sampling approaches.

Conclusions:

• This research bridges network analysis and RDS, enabling the study of social network structures in hidden populations.
• The findings expand the application of RDS, offering new insights into the social factors driving disease and information diffusion.
• This work facilitates a deeper understanding of population structures and their impact on public health outcomes.