Causal inference over stochastic networks.

Area of Science:

• Social network analysis
• Causal inference
• Statistical modeling

Background:

• Causal inference in networks requires addressing outcome dependencies.
• Treatment spillover and outcome interference are critical challenges.
• Existing models often assume network independence or fixed structures.

Purpose of the Study:

• To develop a novel model for causal inference in networks with endogenous structures.
• To jointly model relational and covariate generation processes.
• To overcome limitations of separability and fixed network assumptions.

Main Methods:

• Developed a joint model for endogenous network structures and actor covariates.
• Utilized Exponential-family Random Network models (ERNM).
• Employed a Bayesian framework for potential outcome-based inference and modified the exchange algorithm for sampling.

Main Results:

• The proposed framework successfully models endogenous networks.
• Simulation studies demonstrated the validity of the approach.
• The model effectively handles complex dependencies and spillover effects.

Conclusions:

• The developed framework provides a robust method for causal inference in complex network settings.
• It offers a flexible alternative to restrictive network assumptions.
• The approach is valuable for studying phenomena like adolescent smoking within social networks.