Social Network Mediation Analysis: A Latent Space Approach

Area of Science:

• Social Sciences
• Psychology
• Network Science

Background:

• Social networks are crucial for individual mental health and social development.
• Understanding the mediating role of social connections is essential.
• Existing models may not fully capture the complexity of social network mediation.

Purpose of the Study:

• To propose and validate a novel mediation model where the social network acts as a mediator.
• To account for actor dependence within a latent social space.
• To rigorously define and measure the total mediation effect of a social network.

Main Methods:

• Development of a mediation model incorporating latent social space dimensions.
• Utilizing a Bayesian estimation method for model parameter estimation.
• Extensive simulation studies to evaluate model performance under various conditions.

Main Results:

• Demonstrated that latent dimensions are equivalent and their specific meaning is arbitrary.
• Established that the proposed network mediation effect is well-defined despite non-unique individual positions.
• Validated the model's usefulness through an application to a college friendship network.

Conclusions:

• The proposed network mediation model offers a robust framework for analyzing social network effects.
• This approach provides a quantifiable measure of the total mediation effect of social networks.
• The model has practical applications in understanding social dynamics and their impact on individuals.