A brief primer on conducting regression-based causal mediation analysis.

Abstract

Objective: We provide an overview of regression-based causal mediation analysis in the field of traumatic stress and guidance on how to conduct mediation analysis using our R package regmedint. Method: We discuss the causal interpretations of the quantities that causal mediation analysis estimates, including total, direct, and indirect effects, especially when the interaction between exposure and mediator is permitted. We discuss the assumptions that must be fulfilled for mediation analyses to validly estimate these causal quantities, discuss suitable study designs for assessing mediation, and describe how causal mediation analysis differs from traditional methods of mediation. To illustrate how to conduct and interpret mediation analysis using our R package regmedint, we use data from a published longitudinal study to assess the extent to which children's externalizing behavior mediates changes in parental negative feelings during the COVID-19 lockdown. We compare the results to those obtained using traditional methods, thus illustrating the importance of accounting for exposure–mediator interaction when an interaction may be present. Results: When the exposure and the mediator interact, traditional methods can provide estimates of direct and indirect effects that differ from those provided by more flexible causal mediation methods. When the exposure and the mediator do not interact, traditional methods and causal mediation method may estimate similar direct and indirect effects depending on the model specification. Conclusions: In contrast to traditional methods of mediation analysis, regression-based causal mediation methods seek to estimate specific interventional quantities, not mere associations, and the causal methods explicitly allow for exposure–mediator interactions. We recommend using these methods by default rather than using more restrictive traditional methods.

Effect Measure Modification by Covariates in Mediation: Extending Regression-based Causal Mediation Analysis

“…5 The extended formulas in Table 1 are implemented in the R regmedint package (version 1.0.0 and later). 15–17…”

Introduction to the special section on causal inference and agent-based modeling.

“…Table 1, in the article by Matthay et al (2023), potential sources of instruments in stress and trauma research, may help researchers to identify an appropriate instrument for their own research. Li et al (2023) carefully describe causal mediation analysis and the differences between (and similarities to) traditional mediation analysis, using an example of parental negative feelings and children’s externalizing behavior during pandemic-related lockdown.…”

An introduction to directed acyclic graphs in trauma research.

“…Another way of thinking about a mediator is that it is a variable that has arrows pointing to it (i.e., A → D) and from it (D → Y) on a path between two variables. The total causal effect of A on Y is the direct effect of A on Y (A → Y) as well as the indirect effect of A on Y through the mediator D, that is, A → D → Y (Li et al, 2023; VanderWeele, 2015). Conditioning on D will introduce bias in the estimation of the total effect of A on Y by blocking the indirect effect, that is, A → D → Y.…”

“…Matthay et al (2023) use DAGs to illustrate the difference between confounder control and instrumental variable approaches to causal inference. DAGs are used by Li et al (2023) to illustrate both the structure of a causal mediation analysis as well as a concrete example in which children’s externalizing behavior mediates changes in parental negative feelings during the COVID-19 lockdown.…”

“…A mediator is a variable that is on a causal path between two variables (VanderWeele & Rothman, 2021). In Figure 1, D is a mediator (Li et al, 2023;VanderWeele, 2015). Conditioning on D will introduce bias in the estimation of the total effect of A on Y by blocking the indirect effect, that is, A → D → Y.…”