![[2403.11332] Graph Machine Learning based Doubly Robust Estimator for Network Causal Effects](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2403.11332] Graph Machine Learning based Doubly Robust Estimator for Network Causal Effects
Summary
This paper presents a novel methodology combining graph machine learning and double machine learning to estimate causal effects in social networks, addressing challenges like interference and confounding factors.
Why It Matters
Understanding causal effects in social networks is crucial for various applications, including public health and economics. This research provides a robust framework that enhances the accuracy of causal inference, which is vital for effective decision-making in social sciences.
Key Takeaways
- Introduces a new estimator for causal effects in social networks.
- Addresses challenges of interference and confounding factors effectively.
- Demonstrates semiparametric efficiency under mild regularity conditions.
- Validates the methodology through extensive simulation studies.
- Applies the method to assess the impact of Self-Help Group participation.
Computer Science > Machine Learning arXiv:2403.11332 (cs) [Submitted on 17 Mar 2024 (v1), last revised 19 Feb 2026 (this version, v3)] Title:Graph Machine Learning based Doubly Robust Estimator for Network Causal Effects Authors:Seyedeh Baharan Khatami, Harsh Parikh, Haowei Chen, Sudeepa Roy, Babak Salimi View a PDF of the paper titled Graph Machine Learning based Doubly Robust Estimator for Network Causal Effects, by Seyedeh Baharan Khatami and 4 other authors View PDF Abstract:We address the challenge of inferring causal effects in social network data. This results in challenges due to interference -- where a unit's outcome is affected by neighbors' treatments -- and network-induced confounding factors. While there is extensive literature focusing on estimating causal effects in social network setups, a majority of them make prior assumptions about the form of network-induced confounding mechanisms. Such strong assumptions are rarely likely to hold especially in high-dimensional networks. We propose a novel methodology that combines graph machine learning approaches with the double machine learning framework to enable accurate and efficient estimation of direct and peer effects using a single observational social network. We demonstrate the semiparametric efficiency of our proposed estimator under mild regularity conditions, allowing for consistent uncertainty quantification. We demonstrate that our method is accurate, robust, and scalable via an extensive simulation stu...
