A general approach to causal mediation analysis.
By Kosuke Imai, Luke Keele, Dustin Tingley. 2010
3,722 citations recorded by the source corpus.
Traditionally in the social sciences, causal mediation analysis has been formulated, understood, and implemented within the framework of linear structural equation models. We argue and demonstrate that this is problematic for 3 reasons: the lack of a general definition of causal mediation effects independent of a particular statistical model, the inability to specify the key identification assumption, and the difficulty of extending the framework to nonlinear models. In this article, we propose an alternative approach that overcomes these limitations. Our approach is general because it offers the definition, identification, estimation, and sensitivity analysis of causal mediation effects without reference to any specific statistical model. Further, our approach explicitly links these 4 elements closely together within a single framework. As a result, the proposed framework can accommodate linear and nonlinear relationships, parametric and nonparametric models, continuous and discrete mediators, and various types of outcome variables. The general definition and identification result also allow us to develop sensitivity analysis in the context of commonly used models, which enables applied researchers to formally assess the robustness of their empirical conclusions to violations of the key assumption. We illustrate our approach by applying it to the Job Search Intervention Study. We also offer easy-to-use software that implements all our proposed methods.
Subfields: Experimental and Cognitive Psychology, Management Science and Operations Research, Statistics and Probability, Causal Inference & Discovery
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
In Experimental and Cognitive Psychology
- Judea Pearl
- Stijn Vansteelandt
- Teppei Yamamoto
- S. Utych
- David S. Pedulla
- Tyler J. VanderWeele
- Kristopher J Preacher
- E. T. Tchetgen Tchetgen
In Management Science and Operations Research
- Kristopher J Preacher
- Tyler J. VanderWeele
- Stijn Vansteelandt
- Judea Pearl
- Ilya Shpitser
- E. T. Tchetgen Tchetgen
- Dustin Tingley
- Teppei Yamamoto
In Statistics and Probability
- Kristopher J Preacher
- Tyler J. VanderWeele
- Judea Pearl
- Stijn Vansteelandt
- Ilya Shpitser
- Dustin Tingley
- Kosuke Imai
- E. T. Tchetgen Tchetgen
In Causal Inference & Discovery
- Judea Pearl
- Tyler J. VanderWeele
- Stijn Vansteelandt
- E. T. Tchetgen Tchetgen
- Ilya Shpitser
- Dustin Tingley
- Yonatan Belinkov
- Kosuke Imai
Citations in the corpus
- Causal Mediation Analysis for Interpreting Neural NLP: The Case of Gender Bias (2020, methodology): “The causality literature often focuses on assumptions needed for identification of mediation effects from observed data (Pearl, 2001; Avin et al., 2005; Imai et al., 2010b).”
- The Role of Anger in the Biased Assimilation of Political Information (2018, methodology): “…Baron and Kenny, 1986; MacKinnon, Warsi, and Dwyer, 1995) and allows for sensitivity analyses relevant to the “sequential ignorability assumption” (Imai, Keele, and Tingley, 2010), i.e., the assumption that the researcher has not ignored a phenomenon correlated with both the mediator and the…”
- Advantages of Monte Carlo Confidence Intervals for Indirect Effects (2012, background): “For overviews see Imai, Keele, and Tingley (2010 ), Muthen (2011 ), and Pearl (2010 ).”
- Mediation with Dichotomous Outcomes (2013, background): “The reader might consult Imai, Keele, and Tingley (2010) for details.”
- How Dehumanization Influences Attitudes toward Immigrants (2018, methodology): “…100 on the feeling thermometer, results become a bit stronger, but still do not achieve statistical significance ( = .140, p = .23). xxiii Results are robust to the Average Causal Mediation Effect analysis (ACME) (Imai et al. 2010) on each mediator independently (using a p-value threshold of .10).”
- The Long-Run Impacts of Same-Race Teachers (2017, neutral): “24The mediation analysis is prone to bias caused by what Acharya et al. (2016) call intermediate confounders and what Imai et al. (2010) call the failure of sequential ignorability: the mediators (absences and achievement) are themselves potentially affected by other unobserved mediators.”
- Anxiety, Immigration, and the Search for Information (2014, methodology): “Using the causal mediation approach advocated by Imai, Keele, and Tingley (2010) and Imai, Keele, Tingley, and Yamamoto (2011), Table 3 shows the effect of being in the treatment condition and immigration worries on three measures of information seeking: (1) the 13 We also conducted a 2x2 repeated…”
- On Partial Identification of the Pure Direct Effect (2015, background): “ Taylor et al., 2005; Petersen et al., 2006; Ten Have et al., 2007; Albert, 2008; Goetgeluk et al., 2008; van der Laan and Petersen, 2008; VanderWeele, 2009; VanderWeele and Vansteelandt, 2009, 2010; Imai et al., 2010a,b; Albert and Nelson, 2011; Tchetgen Tchetgen, 2011; VanderWeele, 2011; Albert, 2”
- Advances in mediation analysis: a survey and synthesis of new developments. (2015, background): “In order to lend δ̄(x) (see Equation 5) a causal interpretation, we must assume sequential ignorability (Emsley et al. 2010, Imai et al. 2010c), sometimes termed exchangeability or no unmeasured confounding.”
- A Downward Spiral? Childhood Suspension and the Path to Juvenile Arrest (2018, background): “More formally, Imai et al.’s (2010) framework applies the counterfactual model of causality to the study of mechanisms.”
- The Mediation Myth (2015, background): “Imai, Keele, and Tingley (2010) described methods to evaluate the assumption of no unmeasured confounder of the relation between M and Y.”