Sampling Networks from Their Posterior Predictive Distribution.

Area of Science:

• Social Network Analysis
• Computational Social Science
• Statistical Modeling

Background:

• Leveraging social network knowledge can enhance intervention efficiency.
• Uncertainty in network structure complicates the estimation of intervention effects.
• Accurate network representation is crucial for guiding intervention selection.

Purpose of the Study:

• To develop methods for sampling networks from posterior predictive distributions.
• To address uncertainty in network property estimation for improved simulations.
• To provide guidance on selecting network-based interventions despite structural uncertainty.

Main Methods:

• Sampling networks using posterior predictive distributions.
• Accounting for uncertainty in network property parameters.
• Methods focus on degree distribution, mixing frequency, and clustering.

Main Results:

• Demonstrated network generation methods using simulated data.
• Validated approaches with data from the National Longitudinal Study of Adolescent Health.
• Showcased the importance of capturing uncertainty in network topology.

Conclusions:

• Network sampling from posterior predictive distributions adequately captures parameter uncertainty.
• Accounting for uncertainty is essential for understanding global network topology.
• The proposed methods facilitate more robust network-based intervention design.