Peer Reporting: Sampling Design and Unbiased Estimates.

Area of Science:

• Social Network Analysis
• Statistical Inference
• Computational Social Science

Background:

• Ego-Centric Sampling Method (ECM) estimates population proportions using peer reports, ensuring privacy.
• Conventional ECM is limited by the assumption of homogeneous networks (uniform node degrees).
• Attribute-degree correlations in heterogeneous networks bias traditional ECM estimates.

Purpose of the Study:

• Introduce the Activity Ratio Corrected ECM estimator (ECMac) for unbiased network inference.
• Address limitations of conventional ECM in heterogeneous social networks.
• Develop a privacy-preserving method for accurate population proportion estimation.

Main Methods:

• Recast population proportion estimation into an edge-space formulation using network reciprocity.
• ECMac corrects for dependencies between node degrees and attributes.
• Utilizes only ego-peer data, avoiding the need for full network structure.

Main Results:

• ECMac provides unbiased and stable estimates in heterogeneous networks.
• Demonstrated up to a 70% reduction in estimation error compared to conventional ECM.
• Simulations and real-world network analyses validate ECMac's performance.

Conclusions:

• ECMac offers a theoretically grounded and practically scalable framework for network-based sampling.
• Enhances the reliability of social network analysis in diverse network structures.
• Establishes a robust method for privacy-preserving estimation of population attributes.