causal mediation analysis

This paper addresses the challenge of causal mediation analysis with multiple mediators and exposure-induced confounding. We propose a unified estimation framework based on potential outcomes simulation. Methodologically, we introduce a novel dual-track simulation strategy: “parametric modeling + deep nonparametric approximation.” It accommodates flexible parametric specifications—linear or nonlinear, continuous or discrete mediators—while pioneering the integration of deep neural networks into multivariate causal mediation inference to learn counterfactual distributions and mitigate model misspecification bias. The framework consistently estimates total, direct, and indirect effects; multivariate natural direct and indirect effects; and path-specific effects. Empirical applications to immigration attitudes and preterm birth replicate and extend prior findings, demonstrating the method’s robustness, flexibility, and high-precision estimation capability.

This study addresses the problem of testing whether a treatment effect operates entirely through observed mediators and identifying causal mechanisms under control for covariates. The authors propose a statistical test based on double machine learning, extending— for the first time—the joint evaluation of full mediation and causal mechanism identification to non-randomized treatment settings. By integrating conditional independence testing, the method achieves root-n consistent and asymptotically normal inference even in the presence of high-dimensional covariates. Simulation studies demonstrate favorable finite-sample performance, and the approach is successfully applied to two randomized experiments examining maternal mental health and social norms.

Existing causal mediation analysis relies heavily on untestable multiple ignorability assumptions, undermining the robustness of mechanistic inference. This paper proposes a novel identification framework grounded in heterogeneous treatment effects, which integrates explicit and implicit mediation pathways via causal decomposition—enabling simultaneous identification of total, direct, and indirect effects without requiring multiple ignorability. The method combines flexible heterogeneity modeling, Monte Carlo simulation–based validation, and a dedicated software implementation. It is empirically validated in two real-world applications: public resource governance and voter information dissemination. Simulation studies demonstrate that, compared to prevailing approaches, the proposed method achieves substantially improved estimation accuracy and bias control. By relaxing strong untestable assumptions and offering practical implementation tools, this work advances causal mediation analysis with a more robust, interpretable, and accessible framework for uncovering underlying causal mechanisms.

This paper addresses estimation bias in causal mediation analysis arising from confounding bias in direct and indirect effect estimation. It systematically evaluates estimators for both univariate and multivariate (binary and continuous) mediators. Innovatively, it provides the first unified benchmark evaluation of state-of-the-art estimators—including multiply robust estimators and double machine learning—in multivariate mediation settings, and proposes a comprehensive practical guideline covering identification assumption validation, estimator selection, and implementation. Through extensive simulation studies and empirical analysis using the UK Biobank brain imaging cohort, the methods demonstrate substantial improvements over conventional parametric and nonparametric mediation models across diverse scenarios. Empirical findings reveal that hypertension and obesity exert significant indirect effects on cognitive function predominantly via structural brain changes—particularly reduced gray matter volume—highlighting the critical role of neuroanatomical pathways in cardiometabolic–cognitive associations.

This paper addresses causal mediation analysis under longitudinal continuous interventions with concurrent confounding and mediating variables, focusing on the mechanism through which invasive mechanical ventilation (IMV) affects survival in COVID-19 patients, with acute kidney injury (AKI) as a mediator. Method: Moving beyond static interventions and parametric modeling assumptions, we propose a semiparametric estimation framework grounded in nonparametric structural equation models. It integrates cross-fitted sequential regression with doubly robust pseudo-outcome techniques to achieve efficient, asymptotically normal, and robust estimation. Contribution/Results: We establish novel identification conditions for mediation effects under longitudinal modified treatment policies and uncover the “inconsistent mediation” phenomenon—where direct and indirect effects operate in opposing directions. Applied to real-world clinical data, our method quantifies heterogeneous causal pathways from IMV to survival via AKI, delivering interpretable, high-precision causal evidence for critical care decision-making.

This study addresses the challenge of estimating natural direct and indirect relative risk effects of a binary exposure on a binary outcome in settings involving multiple mediators—continuous, binary, or mixed—under the sequential ignorability assumption. The authors develop a unified regression framework that, for the first time, yields closed-form expressions for causal effects on the relative risk scale while accounting for inter-mediator dependence and interactions between exposure–mediator and mediator–mediator pairs. By operating directly on the relative risk scale, the approach circumvents interpretational difficulties arising from non-collapsibility and enables flexible modeling alongside rigorous uncertainty quantification. Leveraging likelihood-based inference and analytical derivations, the proposed method demonstrates strong validity and practical utility in two empirical applications.

Existing methods struggle to effectively identify and infer causal mediation pathways involving multiple time-varying mediators. This work proposes the first general analytical framework capable of decomposing total effects into path-specific effects, accommodating both continuous and categorical outcomes. The approach innovatively constructs a mediation path testing strategy that rigorously controls Type I error under composite null hypotheses by integrating studentized statistics with data splitting techniques, thereby enabling identifiability and efficient inference in complex mediation chains. Extensive simulations and two large-scale empirical studies demonstrate the method’s marked advantages in estimation accuracy, inferential validity, and statistical power.

This study addresses the limitations of traditional natural direct and indirect effects (NDE/NIE) in causal mediation analysis, which lack skew-symmetry and additivity, often leading to interpretational paradoxes. To overcome these issues, the authors propose cumulative natural direct and indirect effects (CNDE/CNIE), a novel framework that decomposes local causal effects and, under standard identification assumptions, yields valid measures for both continuous and ordinal treatment variables. The proposed CNDE/CNIE satisfy skew-symmetry and additivity, thereby enabling a coherent decomposition of the total effect. Empirical evaluations in linear mediation models—including those with interaction terms—and real-world data demonstrate that CNDE/CNIE provide more robust and interpretable causal estimates compared to conventional NDE/NIE.

This study addresses limitations in existing meta-analytic methods for mediation, which often suffer from confounding bias, missing data, and ill-defined target populations, thereby undermining causal interpretation. To overcome these issues, the authors propose a causally interpretable meta-mediation framework that first transports natural indirect effects from individual studies to a clearly specified target population and leverages studies containing mediator variables—even those not originally designed for mediation analysis—to broaden the evidence base. The framework introduces a novel random-effects model and a nonparametric heterogeneity decomposition method based on ANOVA-type sums of squares, alongside a data-adaptive estimator grounded in semiparametric theory to flexibly accommodate confounding and missingness. Simulation and empirical analyses demonstrate that the proposed approach performs robustly in finite samples and substantially enhances both the causal interpretability and applicability of mediation effect meta-analysis.

本文解决了随机试验中二元中介变量的因果中介分析样本量不足的问题，通过开发NIE和NDE的分析功率和样本量公式来解决。