An Approach to Causal Inference over Stochastic Networks

Duncan A. Clark,Mark S. Handcock

DOI: https://doi.org/10.48550/arXiv.2106.14145

2022-07-15

Abstract:Claiming causal inferences in network settings necessitates careful consideration of the often complex dependency between outcomes for actors. Of particular importance are treatment spillover or outcome interference effects. We consider causal inference when the actors are connected via an underlying network structure. Our key contribution is a model for causality when the underlying network is unobserved and the actor covariates evolve stochastically over time. We develop a joint model for the relational and covariate generating process that avoids restrictive separability assumptions and deterministic network assumptions that do not hold in the majority of social network settings of interest. Our framework utilizes the highly general class of Exponential-family Random Network models (ERNM) of which Markov Random Fields (MRF) and Exponential-family Random Graph models (ERGM) are special cases. We present potential outcome based inference within a Bayesian framework, and propose a simple modification to the exchange algorithm to allow for sampling from ERNM posteriors. We present results of a simulation study demonstrating the validity of the approach. Finally, we demonstrate the value of the framework in a case-study of smoking over time in the context of adolescent friendship networks.

Methodology,Applications

What problem does this paper attempt to address?

This paper attempts to solve the problems encountered in causal inference in network environments, especially the difficult problems of causal inference when the network structure itself is endogenous and randomly co - evolves over time. Specifically, the paper focuses on how to handle the complex dependencies among actors in complex network structures, especially the treatment spillover effect or outcome interference effect. The author proposes a new model to handle this situation, namely a causal inference model when the underlying network structure is endogenous and changes randomly over time.

The key contribution of the paper lies in proposing a joint model for the processes of generating relationships and covariates, avoiding restrictive separation assumptions and deterministic network assumptions, which rarely hold in real - world social network settings. This framework utilizes a highly general class of exponential - family random network models (ERNM), of which Markov random fields (MRF) and exponential - family random graph models (ERGM) are special cases.

By using the Bayesian framework, the paper proposes a causal inference method based on potential outcomes and presents a modified exchange algorithm to allow sampling from the ERNM posterior distribution. In addition, the paper verifies the effectiveness of the method through simulation studies and demonstrates the value of this framework through a case study of smoking behavior over time in adolescent friendship networks.

In summary, this paper aims to address the challenges faced in causal inference in network environments, especially when the network structure itself is dynamic and uncertain, and provides a new modeling and analysis method.