Modeling Psychometric Relational Data in Social Networks: Latent Interdependence Models.

Area of Science:

• Social Network Analysis
• Psychometrics
• Statistical Modeling

Background:

• Traditional social network analysis often overlooks the mutual influence between individuals.
• Social Relations Models (SRMs) offer a framework for understanding mutual-rating processes but can be complex to apply to psychometric data.

Purpose of the Study:

• To propose a novel latent interdependence approach for modeling psychometric data within social networks.
• To introduce two specific psychometric models within this framework: one with main effects and another incorporating a latent distance effect.
• To evaluate the performance of these models using Bayesian estimation and assess their applicability to real-world network data.

Main Methods:

• Development of two psychometric models based on latent interdependence, incorporating sender, receiver, and latent distance effects.
• Utilizing Bayesian estimation via Markov Chain Monte Carlo (MCMC) for parameter estimation.
• Conducting a simulation study to evaluate parameter recovery accuracy and network size effects, followed by analysis of empirical data.

Main Results:

• Both proposed models demonstrated accurate parameter recovery across various conditions in the simulation study.
• Model estimation accuracy significantly improved with increasing network size.
• Empirical data analysis showed the models' utility in predicting latent connection weights and reconstructing network structures at a latent trait level.

Conclusions:

• The latent interdependence approach provides a robust framework for analyzing psychometric data in social networks.
• The proposed models, particularly with larger network sizes, offer accurate insights into relationship dynamics and network structure.
• The methodology facilitates the prediction of connection weights and the rebuilding of networks based on latent characteristics.