Network Causal Effect Estimation In Graphical Models Of Contagion And Latent Confounding

2024-11-02

arXiv.org

AI Summary

This paper addresses the fundamental identifiability challenge in network causal inference: distinguishing whether observed associations arise from contagion effects or latent confounding. Focusing on the universal interference setting—where a single network is realized and units exhibit arbitrary dependence—we propose the *segregated graph*, a unified graphical model encoding both mechanisms. We develop, for the first time, a likelihood-ratio test valid under universal interference, enabling rigorous discrimination between contagion and latent confounding. Building upon this, we design a mechanism-adaptive, unbiased, and consistent estimation framework that relaxes stringent structural assumptions on interference imposed by prior methods. Our theoretical analysis establishes asymptotic statistical properties under network growth. Empirical evaluation confirms the test’s validity on synthetic data and demonstrates the model’s plausibility on real-world social networks. The proposed methodology substantially improves both the accuracy and applicability boundary of causal effect estimation under universal interference.

Technology Category

• Reasoning under Uncertainty: Causality
• Machine Learning: Causal Learning
• Cognitive Modeling & Cognitive Systems: Conceptual Inference and Reasoning

Application Category

• Graph Algorithms and Modeling for the Web: Algorithms and analysis for incomplete, noisy, or partially observed Web-related graphs
• Social Networks and Social Media: Influence propagation, information diffusion, and the prediction on networks
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

A key question in many network studies is whether the observed correlations between units are primarily due to contagion or latent confounding. Here, we study this question using a segregated graph (Shpitser, 2015) representation of these mechanisms, and examine how uncertainty about the true underlying mechanism impacts downstream computation of network causal effects, particularly under full interference -- settings where we only have a single realization of a network and each unit may depend on any other unit in the network. Under certain assumptions about asymptotic growth of the network, we derive likelihood ratio tests that can be used to identify whether different sets of variables -- confounders, treatments, and outcomes -- across units exhibit dependence due to contagion or latent confounding. We then propose network causal effect estimation strategies that provide unbiased and consistent estimates if the dependence mechanisms are either known or correctly inferred using our proposed tests. Together, the proposed methods allow network effect estimation in a wider range of full interference scenarios that have not been considered in prior work. We evaluate the effectiveness of our methods with synthetic data and the validity of our assumptions using real-world networks.

Problem

• Distinguish contagion vs latent confounding in network correlations.
• Estimate causal effects under full interference in networks.
• Develop tests and strategies for unbiased network effect estimation.

Innovation

• Uses segregated graph for contagion and confounding analysis
• Develops likelihood ratio tests for dependence mechanisms
• Proposes unbiased network causal effect estimation strategies