🤖 AI Summary
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.
📝 Abstract
Analyses of causal mediation often involve exposure-induced confounders or, relatedly, multiple mediators. In such applications, researchers aim to estimate a variety of different quantities, including interventional direct and indirect effects, multivariate natural direct and indirect effects, and/or path-specific effects. This study introduces a general approach to estimating all these quantities by simulating potential outcomes from a series of distribution models for each mediator and the outcome. Building on similar methods developed for analyses with only a single mediator (Imai et al. 2010), we first outline how to implement this approach with parametric models. The parametric implementation can accommodate linear and nonlinear relationships, both continuous and discrete mediators, and many different types of outcomes. However, it depends on correct specification of each model used to simulate the potential outcomes. To address the risk of misspecification, we also introduce an alternative implementation using a novel class of nonparametric models, which leverage deep neural networks to approximate the relevant distributions without relying on strict assumptions about functional form. We illustrate both methods by reanalyzing the effects of media framing on attitudes toward immigration (Brader et al. 2008) and the effects of prenatal care on preterm birth (VanderWeele et al. 2014).