Bounds for causal mediation effects

2025-12-12

AI Summary

This paper addresses the identification of sharp bounds for causal mediation effects under unobserved confounding, focusing on two dominant frameworks: natural direct/indirect effects and separable effects. Methodologically, it integrates the potential outcomes framework, symbolic bound analysis, and sensitivity analysis to characterize bounds when cross-world independence fails. The key contribution is the first rigorous proof that, under randomized treatment assignment but with unobserved confounding affecting the mediator, the sharp symbolic bounds derived from both frameworks are identical. It further systematically evaluates bound tightness and robustness under violations of cross-world independence. Applied to a clinical trial on peanut allergy, the proposed bounds are substantially narrower than conventional conservative estimates, enabling more precise quantification of the mediation effect through immunological biomarker pathways. This advancement enhances mechanistic interpretability and strengthens the practical utility of causal mediation analysis in settings with unmeasured confounding.

Technology Category

• Reasoning under Uncertainty: Causality
• Machine Learning: Causal Learning
• Knowledge Representation and Reasoning: Action, Change, and Causality

Application Category

• Graph Algorithms and Modeling for the Web: Algorithms and analysis for incomplete, noisy, or partially observed Web-related graphs
• Responsible Web: Human-perceived consequences of algorithmic deployment on the web
• Social Networks and Social Media: Analysis of web-mediated communications, communities, and crowds

Abstract

Several frameworks have been proposed for studying causal mediation analysis. What these frameworks have in common is that they all make assumptions for point identifications that can be violated even when treatment is randomized. When a causal effect is not point-identified, one can sometimes derive bounds, i.e. a range of possible values that are consistent with the observed data. In this work, we study causal bounds for mediation effects under both the natural effects framework and the separable effects framework. In particular, we show that when there are unmeasured confounders for the intermediate variables(s) the sharp symbolic bounds on separable (in)direct effect coincide with existing bounds for natural (in)direct effects in the analogous setting. We compare these bounds to valid bounds for the natural direct effects when only the cross-world independence assumption does not hold. Furthermore, we demonstrate the use and compare the results of the bounds on data from a trial investigating the effect of peanut consumption on the development of peanut allergy in infants through specific pathways of measured immunological biomarkers.

Problem

• Derives bounds for causal mediation effects under unmeasured confounding
• Compares separable and natural effects frameworks in mediation analysis
• Applies bounds to peanut allergy trial data for pathway analysis

Innovation

• Bounds for causal mediation effects under unmeasured confounders
• Comparison of natural and separable effects frameworks
• Application to peanut allergy trial with immunological biomarkers