Average Controlled and Average Natural Micro Direct Effects in Summary Causal Graphs

2024-10-31

arXiv.org

AI Summary

This study addresses the identifiability of the Average Controlled Direct Effect (ACDE) and Average Natural Direct Effect (ANDE) in abstract causal graphs—dynamic system abstractions characterized by cycles and missing temporal information—under realistic, nonlinear epidemiological settings with latent confounding. We establish the first sufficient identifiability conditions for micro-level direct effects within such abstract causal graphs. We prove that these conditions are also necessary under the assumptions of no latent confounding and identification via adjustment alone. Departing from conventional linear or parametric assumptions, our framework supports nonparametric definitions of causal effects. Methodologically, we integrate an extended do-calculus, adjustment-based identification, and latent-confounder analysis. Our results yield rigorous, graph-theoretic identifiability criteria, substantially broadening both the theoretical scope and practical applicability of direct effect estimation in dynamic, complex systems.

Technology Category

• Reasoning under Uncertainty: Causality
• Machine Learning: Causal Learning
• Knowledge Representation and Reasoning: Diagnosis and Abductive Reasoning

Application Category

• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Responsible Web: Human-perceived consequences of algorithmic deployment on the web
• Economics, Online Markets and Human Computation: Incentives in network design for Web infrastructures and ecosystems

Abstract

In this paper, we investigate the identifiability of average controlled direct effects and average natural direct effects in causal systems represented by summary causal graphs, which are abstractions of full causal graphs, often used in dynamic systems where cycles and omitted temporal information complicate causal inference. Unlike in the traditional linear setting, where direct effects are typically easier to identify and estimate, non-parametric direct effects, which are crucial for handling real-world complexities, particularly in epidemiological contexts where relationships between variables (e.g, genetic, environmental, and behavioral factors) are often non-linear, are much harder to define and identify. In particular, we give sufficient conditions for identifying average controlled micro direct effect and average natural micro direct effect from summary causal graphs in the presence of hidden confounding. Furthermore, we show that the conditions given for the average controlled micro direct effect become also necessary in the setting where there is no hidden confounding and where we are only interested in identifiability by adjustment.

Problem

• Identifying direct effects in non-parametric causal graphs with hidden confounding
• Providing sufficient conditions for controlled and natural micro direct effects
• Establishing necessary conditions for identifiability via adjustment methods

Innovation

• Identifies direct effects in summary causal graphs
• Handles hidden confounding in non-parametric settings
• Provides necessary conditions for adjustment identifiability