Interpretation and identification of causal mediation
Open accessThe DOI matches its Crossref record; 88 of the 88 references it cites were found in a registry or an index.
Abstract
This article reviews the foundations of causal mediation analysis and offers a general and transparent account of the conditions necessary for the identification of natural direct and indirect effects, thus facilitating a more informed judgment of the plausibility of these conditions in specific applications. I show that the conditions usually cited in the literature are overly restrictive and can be relaxed substantially without compromising identification. In particular, I show that natural effects can be identified by methods that go beyond standard adjustment for confounders, applicable to observational studies in which treatment assignment remains confounded with the mediator or with the outcome. These identification conditions can be validated algorithmically from the diagrammatic description of one's model and are guaranteed to produce unbiased results whenever the description is correct. The identification conditions can be further relaxed in parametric models, possibly including interactions, and permit one to compare the relative importance of several pathways, mediated by interdependent variables.
Overview
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
Citations
Cites
Baron, R. M., & Kenny, D. A. (1986) · Journal of Personality and Social Psychology · doi:10.1037/0022-3514.51.6.1173 · Cited by 91,532
The central role of the propensity score in observational studies for causal effects
ROSENBAUM, P. R., & RUBIN, D. B. (1983) · Biometrika · doi:10.1093/biomet/70.1.41 · Cited by 31,241
Structural Equations with Latent Variables.
Clogg, C. C., & Bollen, K. A. (1991) · Contemporary Sociology · doi:10.2307/2072165 · Cited by 17,442
Causality: Models, Reasoning, and Inference
Pearl, J. (2009) · doi:10.1017/cbo9780511803161 · Cited by 11,653
Self-citationAddressing Moderated Mediation Hypotheses: Theory, Methods, and Prescriptions
Preacher, K. J., Rucker, D. D., & Hayes, A. F. (2007) · Multivariate Behavioral Research · doi:10.1080/00273170701341316 · Cited by 9,973
Beyond Baron and Kenny: Statistical Mediation Analysis in the New Millennium
Hayes, A. F. (2009) · Communication Monographs · doi:10.1080/03637750903310360 · Cited by 9,757 · cited 4×
“, Imai, Keele, & Yamamoto, 2010) or uncorrelated error terms. 3 Imai, Keele, and Yamamoto (2010) and Imai, Keele, Tingley, and Yamamoto (2011) discussed similarities and differences among several versions of the identifying assumptions, and Shpitser and VanderWeele (2011) delineated the context under which a restricte…”
InfluentialEstimating causal effects of treatments in randomized and nonrandomized studies
Rubin, D. B. (1974) · Journal of Educational Psychology · doi:10.1037/h0037350 · Cited by 9,436 · cited 2×
“Counterfactual thinking in statistics (Holland, 1988; Rubin, 1974) and epidemiology (Robins and Greenland, 1992), together with a formal semantics based on non-parametric structural equations (Balke and Pearl, 1995; Halpern, 1998; Pearl, 2001) have given causal mediation analysis a sound theoretical basis and have ext…”
Introduction to Statistical Mediation Analysis
MacKinnon, D. P. (2012) · doi:10.4324/9780203809556 · Cited by 5,618
Fisher, R. A. (1936) · BMJ · doi:10.1136/bmj.1.3923.554-a · Cited by 4,980
CAUSALITY: MODELS, REASONING AND INFERENCE
Hitchcock, C. (2001) · The Philosophical Review · doi:10.1215/00318108-110-4-639 · Cited by 4,741
Self-citation
Cited by
Items 101–200 of 554
Causal links among amyloid, tau, and neurodegeneration
Bilgel, M., Wong, D. F., Moghekar, A. R., Ferrucci, L., Resnick, S. M., & Initiative, T. A. D. N. (2022) · Brain Communications · doi:10.1093/braincomms/fcac193 · Cited by 40
“41 The average causal mediation effect (ACME; also known as the natural indirect effect) for the amyloid negative (A–) group, denoted ACME A– , is the expected difference in the outcome while amyloid status is fixed at A– and the mediator (tau) changes from the value attained or that would have…”
Loh, W. W., & Ren, D. (2023) · Perspectives on Psychological Science · doi:10.1177/17456916221134573 · Cited by 45
“The assumption of no unmeasured confounding (also termed “exchangeability” or “unconfoundedness”) of the mediator–outcome relation has been formally stated in the literature as a component of the “sequential ignorability” assumptions (Imai, Keele, & Tingley, 2010; Imai, Keele, & Yamamoto, 2010; Pearl, 2014; Shpitser &…”
BackgroundComment on Pearl: Practical implications of theoretical results for causal mediation analysis.
Imai, K., Keele, L. J., Tingley, D. H., & Yamamoto, T. (2014) · Psychological Methods · doi:10.1037/met0000021 · Cited by 49
Côté, S. M., Ahun, M. N., Herba, C. M., Brendgen, M. R., Geoffroy, M., Orri, M., Liu, X., Vitaro, F., Melchior, M., Boivin, M., et al. (2018) · Journal of the American Academy of Child & Adolescent Psychiatry · doi:10.1016/j.jaac.2018.04.024 · Cited by 50
Probabilistic Computational Causal Discovery for Systems Biology
Lagani, V., Triantafillou, S., Ball, G., Tegner, J. N., & Tsamardinos, I. (2015) · Studies in mechanobiology, tissue engineering and biomaterials · doi:10.1007/978-3-319-21296-8_3 · Cited by 46
“However, these algorithms usually assume some particular (l inear) distributional model and “fell short of providing a general, causally defensible measure of mediation” [114].”
BackgroundCausal Interventional Training for Image Recognition
Qin, W., Zhang, H., Hong, R., Lim, E., & Sun, Q. (2021) · IEEE Transactions on Multimedia · doi:10.1109/tmm.2021.3136717 · Cited by 37
Romo, M. L., George, G., Mantell, J. E., Mwai, E., Nyaga, E. N., Strauss, M., Odhiambo, J. O., Govender, K., & Kelvin, E. A. (2019) · PeerJ · doi:10.7717/peerj.7253 · Cited by 34
“This analysis was conducted using a SAS Macro developed by Valeri & VanderWeele (2013), with mediation methodology from Pearl (2014).”
MethodCausal inference in cumulative risk assessment: The roles of directed acyclic graphs
Riess, L. E., Wright, J. M., Rice, G. E., Neas, L., & Teuschler, L. K. (2016) · Environment International · doi:10.1016/j.envint.2016.12.005 · Cited by 40
“Appendix C provides further explanation of how analysis of direct and indirect effects can generate casual inferences for dose-response assessment in CRAs (Pearl, 2014; Richiardi et al., 2013).”
BackgroundThe Varying Effects of Incarceration, Conviction, and Arrest on Wealth Outcomes among Young Adults
Maroto, M. L., & Sykes, B. L. (2019) · Social Problems · doi:10.1093/socpro/spz023 · Cited by 33
Loh, W. W., Moerkerke, B., Loeys, T., & Vansteelandt, S. (2021) · Psychological Methods · doi:10.1037/met0000314 · Cited by 46
Belief in COVID-19 Conspiracy Theories Reduces Social Distancing over TimePreprint
Bierwiaczonek, K., Kunst, J. R., & Pich, O. (2020) · doi:10.31234/osf.io/tqfrw · Cited by 35
Henry, A. C., van Dongen, J. C., van Goor, I. W. J. M., Smits, F. J., Nagelhout, A., Besselink, M. G., Busch, O. R., Bonsing, B. A., Bosscha, K. P., van Dam, R. M., et al. (2023) · BJS Open · doi:10.1093/bjsopen/zrac174 · Cited by 40
“0 542 (51) 552 (52) Major complications‡ 306 (29) 2 (0) – Length of hospital stay (days), median (i.”
Mechanisms and mediation in survival analysis: towards an integrated analytical framework
Pratschke, J., Haase, T., Comber, H., Sharp, L., Cancela, M. D. C., & Johnson, H. (2016) · BMC Medical Research Methodology · doi:10.1186/s12874-016-0130-6 · Cited by 40
“The direct effects in the statistical model are equivalent to “natural direct effects”, whilst the total effect is given by the sum of the direct and indirect effects [10].”
King, T. L., Aitken, Z., Milner, A. J., Emerson, E., Priest, N., Karahalios, A., Kavanagh, A. M., & Blakely, T. A. (2018) · International Journal of Epidemiology · doi:10.1093/ije/dyy154 · Cited by 38
Milovanović, I., Matic, R. M., Alexandris, K., Maksimović, N., Milošević, Z., & Drid, P. (2019) · Journal of Hospitality & Tourism Research · doi:10.1177/1096348019883920 · Cited by 55
“The mediation analysis (Judd & Kenny, 2010; Pearl, 2014) was tested using Sobel test (p < .001). results The findings of the research are classified in three parts: (1) descriptive statistics, (2) confirmatory factor analysis, and (3) structural equation modeling.”
MethodVerostek, M., Miller, C. W., & Zwickl, B. (2021) · Physical Review Physics Education Research · doi:10.1103/physrevphyseducres.17.020115 · Cited by 31
“Although the assumptions prescribed by causal mediation analysis may be stated in multiple ways [72, 73], we describe them in the manner presented by [18], who condenses them into four requirements.”
Bias mitigation in empirical peace and conflict studies: A short primer on posttreatment variables
Dworschak, C. (2023) · Journal of Peace Research · doi:10.1177/00223433221145531 · Cited by 42
“Depending on functional form and model setup, more bias accumulates quickly (Imai, Keele & Tingley, 2010; Glynn, 2012; Pearl, 2014).”
BackgroundMethodInfluentialThoresen, S., Aakvaag, H. F., Strøm, I. F., Wentzel‐Larsen, T., & Birkeland, M. S. (2018) · Social Science & Medicine · doi:10.1016/j.socscimed.2018.06.002 · Cited by 38
Ben‐Eliyahu, A., & Zeidner, M. (2020) · Cambridge University Press eBooks · doi:10.1017/9781108264822.041 · Cited by 34
Comment: Understanding Simpson’s Paradox
Pearl, J. (2022) · ACM eBooks · doi:10.1145/3501714.3501738 · Cited by 56
Self-citationLindmark, A., de Luna, X., & Eriksson, M. (2018) · Statistics in Medicine · doi:10.1002/sim.7620 · Cited by 37
“E (Yi|Zi = 1,Mi = m,X i = x) in (4) with E (Yi|Zi = 0,Mi = m,X i = x).”
BackgroundMethodInfluentialWright, D. B., Eaton, A. A., & Skagerberg, E. M. (2014) · Journal of Research in Personality · doi:10.1016/j.jrp.2014.06.004 · Cited by 37
An Introduction to Causal Mediation Analysis With a Comparison of 2 R Packages
Byeon, S., & Lee, W. (2023) · Journal of Preventive Medicine and Public Health · doi:10.3961/jpmph.23.189 · Cited by 38
“Moreover, Pearl [28,29] suggested the implications of the cross-world independence assumption and related al - ternatives under mild conditions in terms of the non-paramet - ric structural equation models.”
BackgroundEnhancing Out-of-distribution Generalization on Graphs via Causal Attention Learning
Sui, Y., Mao, W., Wang, S., Wang, X., Wu, J., He, X., & Chua, T. (2024) · ACM Transactions on Knowledge Discovery from Data · doi:10.1145/3644392 · Cited by 35
“Equation (5) is usually called backdoor adjustment [53], which is a powerful tool to eliminate the confounding effect.”
BackgroundMethodInfluentialFibre & fermented foods: differential effects on the microbiota-gut-brain axis
Schneider, E., Balasubramanian, R., Ferri, A., Cotter, P. D., Clarke, G. A., & Cryan, J. F. (2024) · Proceedings of The Nutrition Society · doi:10.1017/s0029665124004907 · Cited by 51
Genetic nurturing, missing heritability, and causal analysis in genetic statistics
Shen, H., & Feldman, M. W. (2020) · Proceedings of the National Academy of Sciences · doi:10.1073/pnas.2015869117 · Cited by 28
A Bayesian model selection approach to mediation analysis
Crouse, W. L., Keele, G. R., Gastonguay, M. S., Churchill, G. A., & Valdar, W. S. J. (2022) · PLoS Genetics · doi:10.1371/journal.pgen.1010184 · Cited by 38
“Mediation analysis requires that X, M, and Y are measured in the same individuals, and its causal interpretation relies on assumptions that are not verifiable based solely on the data [1, 7, 8].”
BackgroundOn Partial Versus Full Mediation and the Importance of Effect Sizes
Ledermann, T., Rudaz, M., & Fritz, M. S. (2025) · Advances in Methods and Practices in Psychological Science · doi:10.1177/25152459251355585 · Cited by 51
“First, no statistical effect is an unbiased estimate of a true effect because of the omission of putative confounders or mediators or violations of the assumptions underlying the mediation analysis (e.g., MacKinnon, 2008; MacKinnon & Pirlott, 2015; Pearl, 2014; VanderWeele, 2015).”
Meneton, P., Hoertel, N., Wiernik, E., Lemogne, C., Ribet, C., Bonenfant, S., Ménard, J., Goldberg, M., & Zins, M. (2018) · Social Science & Medicine · doi:10.1016/j.socscimed.2018.09.042 · Cited by 33
Sepriano, A. R., Ramiro, S., van der Heijde, D. M. F. M., et al. (2021) · RMD Open · doi:10.1136/rmdopen-2021-001654 · Cited by 29
Pachankis, J. E., Clark, K. A., Klein, D. N., & Dougherty, L. R. (2021) · Journal of Youth and Adolescence · doi:10.1007/s10964-021-01532-x · Cited by 31
“Finally, mediation analyses were conducted within a potential outcomes framework using PROC CAUSALMED (Valeri & VanderWeele, 2013) to calculate the total, direct, and indirect effects of sexual orientation on internalizing symptoms via social stressors (Pearl, 2014; Valeri & VanderWeele, 2013; Vanderweele & Vansteelan…”
MethodRobust Emotion Recognition in Context Debiasing
Yang, D., Yang, K., Li, M., Wang, S., Wang, S., & Zhang, L. (2024) · IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings · doi:10.1109/cvpr52733.2024.01183 · Cited by 59
“As shown in Figure 3b, when P is set to p ∗ , and the descendant M is changed, we have According to the causal theory [38], The Total Effect (TE) of treatment P = p on Q by comparing the two hypothetical outcomes is formulated as: TE can be disentangled into the Natural Direct Effect (NDE) and the…”
Exploratory Mediation Analysis of Dichotomous Outcomes via Regularization
Serang, S., & Jacobucci, R. (2019) · Multivariate Behavioral Research · doi:10.1080/00273171.2019.1608145 · Cited by 28
“One popular approach is causal mediation analysis (Daniel, De Stavola, Cousens, & Vansteelandt, 2015; VanderWeele & Vansteelandt, 2014; see also Imai, Keele, & Tingley, 2010; Pearl, 2014).”
BackgroundMissing data, imputation, and endogeneity
McDonough, I. K., & Millimet, D. L. (2017) · Journal of Econometrics · doi:10.1016/j.jeconom.2017.05.006 · Cited by 27
“…expenditure per pupil on public primary and secondary school.10 The set of controls is intentionally parsimonious as we do not wish to hold constant current attributes of the children that may act as mediators along the causal pathway between birth weight and current cognitive ability (Pearl 2014).”
BackgroundTahmassebi, A., Motamedi, M., Alavi, A. H., & Gandomi, A. H. (2021) · Engineering Computations · doi:10.1108/ec-02-2021-0096 · Cited by 27
Cankaya, B., Topuz, K., Delen, D., & Glassman, A. (2023) · Omega · doi:10.1016/j.omega.2023.102906 · Cited by 35
“started to make bottom-up and top-down reasoning to find answers to research questions [48] .”
BackgroundSensitivity Analysis of the No-Omitted Confounder Assumption in Latent Growth Curve Mediation Models
Tofighi, D., Hsiao, Y., Kruger, E. S., MacKinnon, D. P., Horn, M. L. V., & Witkiewitz, K. A. (2018) · Structural Equation Modeling A Multidisciplinary Journal · doi:10.1080/10705511.2018.1506925 · Cited by 31
“…the assumption of no omitted confounders, which is necessary to produce unbiased, causally interpretable estimates of the indirect effect (Imai, Keele, & Tingley, 2010; Judd & Kenny, 1981; MacKinnon & Pirlott, 2015; Pearl, 2014; Robins & Greenland, 1992; Valente, Pelham, Smyth, &MacKinnon, 2017).”
BackgroundDeYoung, C., & Blain, S. D. (2020) · The Cambridge Handbook of Personality Psychology · doi:10.1017/9781108264822.026 · Cited by 16
Identification of Causal Mechanisms from Randomized Experiments
Peng, J. (2019) · Information Systems Research · doi:10.1287/isre.2022.1113 · Cited by 28
Gao, M., Roy, A., Deluty, A., Sharkey, K. M., Hoge, E. A., Liu, T., & Brewer, J. A. (2022) · Psychosomatic Medicine · doi:10.1097/psy.0000000000001083 · Cited by 24
“In the causal mediation analysis, we focused on the natural direct and indirect effects and used the bootstrap method with 1000 bootstrapped resamples to compute the standard errors and 95% confidence interval (CI) (39,40).”
MethodWellbeing at home: a mediation analysis of residential satisfaction, comfort, and home attachment
Wegener, B. A., & Schmidt, P. (2023) · Journal of Housing and the Built Environment · doi:10.1007/s10901-023-10068-4 · Cited by 39
“In view of these theoretical and practical concerns, this paper addresses the causal relationship of comfort sensations, home attachment and residential satisfaction in an empirical mediation analysis (Pearl, 2014).”
CWGCNA: an R package to perform causal inference from the WGCNA framework
Liu, Y. (2024) · NAR Genomics and Bioinformatics · doi:10.1093/nargab/lqae042 · Cited by 31
“Besides, it has applications in environmental and epidemiological studies, such as the observation that 8-isoprostane, an oxidative stress biomarker, mediates the effect of plasticizer exposure on preterm birth ( 9 ).”
BackgroundThe equigenic potential of green window views for city dwellers’ well-being
Kley, S. A., & Dovbischuk, T. (2024) · Sustainable Cities and Society · doi:10.1016/j.scs.2024.105511 · Cited by 30
Spillover Effects in Experimental Data
Aronow, P. M., Eckles, D., Samii, C., & Zonszein, S. (2021) · Cambridge University Press eBooks · doi:10.1017/9781108777919.021 · Cited by 44
“Identifying mediation effects requires that other types of conditional independence hold, such as sequential ignorabilty (Imai, Keele and Yamamoto, 2010; Pearl, 2014).”
BackgroundCausality validation of multilevel flow modelling
Nielsen, E. K., Gofuku, A., Zhang, X., et al. (2020) · Computers & Chemical Engineering · doi:10.1016/j.compchemeng.2020.106944 · Cited by 23
Reisenzein, R., Hildebrandt, A., & Weber, H. (2020) · The Cambridge Handbook of Personality Psychology · doi:10.1017/9781108264822.009 · Cited by 1
Distilling Causal Effect from Miscellaneous Other-Class for Continual Named Entity Recognition
Zheng, J., Liang, Z., Chen, H., & Ma, Q. (2022) · Conference on Empirical Methods in Natural Language Processing (EMNLP) · doi:10.18653/v1/2022.emnlp-main.236 · Cited by 46
Nuzzo, J. L., Finn, H. T., & Herbert, R. (2019) · Sports Medicine · doi:10.1007/s40279-019-01131-8 · Cited by 28
Nonparametric inference for interventional effects with multiple mediators
Benkeser, D. C., & Ran, J. (2021) · Journal of Causal Inference · doi:10.1515/jci-2020-0018 · Cited by 23
“…for posing interesting scientific questions pertaining to the mediation of effects through specific biologic pathways (among others, Imai et al. (2010); Valeri and VanderWeele (2013); Pearl (2014); Naimi et al. (2016); Zheng and van der Laan (2017); VanderWeele and Tchetgen Tchetgen (2017)).”
BackgroundTafti, A., & Shmueli, G. (2019) · Information Systems Research · doi:10.1287/isre.2020.0938 · Cited by 30
Chadha, A. R., Arora, G., & Kaloty, N. (2020) · arXiv (Cornell University) · doi:10.48550/arxiv.2011.07735 · Cited by 42
“reasoning [43, 42].”
How Does Counterfactually Augmented Data Impact Models for Social Computing Constructs?
Sen, I., Samory, M., Flöck, F., Wagner, C., & Augenstein, I. (2021) · Conference on Empirical Methods in Natural Language Processing (EMNLP) · doi:10.18653/v1/2021.emnlp-main.28 · Cited by 39
“mediation (Pearl, 2014), we put forth a typology of construct-driven and construct-agnostic CAD.”
MethodYang, D., Yang, K., Kuang, H., Chen, Z., Wang, Y., & Zhang, L. (2024) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2024.3443129 · Cited by 34
“This philosophy can be viewed as conducting a randomized controlled experiment by collecting images of subjects with any emotion in any context.”
BackgroundSocial Network Mediation Analysis: A Latent Space Approach
Liu, H., Jin, I. H., Zhang, Z. J., & Yuan, Y. (2020) · Psychometrika · doi:10.1007/s11336-020-09736-z · Cited by 39
“To draw causal conclusions using the regression approach, one must know that it is a causal path from X to M to Y based on theory (Imai et al., 2010; Pearl, 2014).”
BackgroundTask-relevant and task-irrelevant variability causally shape error-based motor learning
Dal’Bello, L. R., & Izawa, J. (2021) · Neural Networks · doi:10.1016/j.neunet.2021.07.015 · Cited by 21
González‐Serrano, M. H., Valantinė, I., Matic, R. M., Milovanović, I., Сушко, Р., & Moreno, F. C. (2023) · European Research on Management and Business Economics · doi:10.1016/j.iedeen.2023.100229 · Cited by 34
CaMIL: Causal Multiple Instance Learning for Whole Slide Image Classification
Chen, K., Sun, S., & Zhao, J. (2024) · Proceedings of the AAAI Conference on Artificial Intelligence · doi:10.1609/aaai.v38i2.27873 · Cited by 54
“…incorporated causal inference into various machine learning tasks, such as semantic segmentation (Zhang et al. 2020), visual categorization (Rao et al. 2021), vision-language tasks (Yang et al. 2021; Niu et al. 2021; Wang et al. 2020), and reinforcement learning (Everitt et al. 2021; Pearl 2014).”
BackgroundNielsen, R. O., Bertelsen, M. L., Møller, M., Hulme, A., Mansournia, M. A., Casals, M., & Parner, E. T. (2020) · British Journal of Sports Medicine · doi:10.1136/bjsports-2018-100245 · Cited by 19
Jaccard, J. J., & Bo, A. (2018) · Journal of the Society for Social Work and Research · doi:10.1086/701388 · Cited by 20
“For other strategies for dealing with omitted variables, see Pearl (2014) and B. O. Muthén, Muthén, and Asparouhov (2016).”
BackgroundBrave the Wind and the Waves: Discovering Robust and Generalizable Graph Lottery Tickets
Wang, K., Liang, Y., Li, X., Li, G., Ghanem, B., Zimmermann, R., Zhou, Z., Yi, H., Zhang, Y., & Wang, Y. (2023) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2023.3342184 · Cited by 42
Rondanelli, M., Perna, S., Alalwan, T. A., Cazzola, R., Gasparri, C., Infantino, V., Perdoni, F., Iannello, G., Pepe, D., & Guido, D. (2020) · Lipids in Health and Disease · doi:10.1186/s12944-020-01308-5 · Cited by 19
Mustieles, V., Pérez-Carrascosa, F. M., León, J., Lange, T., Bonde, J., Gómez-Peña, C., Artacho‐Cordón, F., Barrios‐Rodríguez, R., Olmedo‐Requena, R., Expósito, J., et al. (2021) · Environmental Science & Technology · doi:10.1021/acs.est.0c08180 · Cited by 25
Kemp, S., & Gómez, D. V. (2022) · The British Journal of Criminology · doi:10.1093/bjc/azac068 · Cited by 23
Reply to commentary by Imai, Keele, Tingley, and Yamamoto concerning causal mediation analysis.
Pearl, J. (2014) · Psychological Methods · doi:10.1037/met0000022 · Cited by 19
Self-citationCausal mediation effects in single case experimental designs.
Valente, M. J., Rijnhart, J. J. M., & Miočević, M. (2023) · Psychological Methods · doi:10.1037/met0000497 · Cited by 21
Loh, W. W., & Ren, D. (2022) · Social and Personality Psychology Compass · doi:10.1111/spc3.12708 · Cited by 22
Hutton, J., Puyat, J. H., Asamoah-Boaheng, M., Sobolev, B., Lingawi, S., Khalili, M., Kuo, C., Shadgan, B., Christenson, J. M., & Grunau, B. E. (2023) · Resuscitation · doi:10.1016/j.resuscitation.2023.109906 · Cited by 23
Chest X-ray Image Classification: A Causal Perspective
Nie, W., Zhang, C., Song, D., Bai, Y., Xie, K., & Liu, A.-A. (2023) · Lecture notes in computer science · doi:10.1007/978-3-031-43898-1_3 · Cited by 24
Directionality of Effects in Causal Mediation Analysis
Wiedermann, W., & von Eye, A. (2016) · Wiley series in probability and statistics · doi:10.1002/9781118947074.ch4 · Cited by 21
Testing mediation effects in cross-classified multilevel data
Luo, W. (2016) · Behavior Research Methods · doi:10.3758/s13428-016-0723-3 · Cited by 20
“As was pointed out by Pearl (2014), this approach has two major drawbacks when it is used to estimate causal mediation.”
BackgroundMethodDistilling Causal Effect of Data in Class-Incremental LearningPreprint
Hu, X., Tang, K., Miao, C., Hua, X., & Zhang, H. (2021) · arXiv (Cornell University) · doi:10.48550/arxiv.2103.01737 · Cited by 16
“Based on the graph, we can formulate the causal effect between variables by causal intervention, which is denoted as the do(·) operation [30, 31].”
MethodBendtsen, M., Thomas, K., Linderoth, C., & Bendtsen, P. (2020) · JMIR Research Protocols · doi:10.2196/23677 · Cited by 16
“For the mediator outcomes, mediators will be explored using a causal inference framework [32-34] using Bayesian inference to estimate the natural direct effect and natural indirect effect (as per the definitions of Pearl [34]).”
Roles of C-reactive protein on the association between urinary cadmium and type 2 diabetes
Xiao, L., Zhou, Y., Ma, J., Cao, L., Zhu, C., Li, W., Wang, D., Fan, L., Ye, Z., & Chen, W. (2019) · Environmental Pollution · doi:10.1016/j.envpol.2019.113341 · Cited by 15
Pelly, D., Daly, M., Delaney, L. D., & Doyle, O. (2022) · Frontiers in Psychology · doi:10.3389/fpsyg.2022.823080 · Cited by 32
“In addition, many of the time-varying variables (e.g., income or physical health) are potential mechanisms or outcomes of the COVID-19 restrictions in their own right, therefore it is not appropriate to control for them in the wellbeing equations (cf. Pearl, 1999, 2014).”
BackgroundGuido, D., Perna, S., Peroni, G., Guerriero, F., & Rondanelli, M. (2015) · Aging Clinical and Experimental Research · doi:10.1007/s40520-015-0360-1 · Cited by 17
Confounder detection in linear mediation models: Performance of kernel-based tests of independence
Wiedermann, W., & Li, X. (2019) · Behavior Research Methods · doi:10.3758/s13428-019-01230-4 · Cited by 18
“…for the direct and indirect effects (cf., e.g., Imai, Keele, Tingley, & Yamamoto, 2014; Loeys, Talloen, Goubert, Moerkerke, & Vansteelandt, 2016; Pearl, 2014): (A1) no unmeasured confounder of the relation between x and m, (A2) no unmeasured confounder of the relation between x and y, (A3)…”
BackgroundDirected Digital Dissidence in Autocracies: How China Wins Online
Gainous, J., Han, R., MacDonald, A. W., & Wagner, K. M. (2023) · doi:10.1093/oso/9780197680384.001.0001 · Cited by 20
Brunton, N. M., Dufault, B., Dart, A. B., Azad, M. B., & McGavock, J. (2021) · International Journal of Obesity · doi:10.1038/s41366-021-00930-2 · Cited by 19
The Role of Parenthood for Life Satisfaction of Older Women and Men in Europe
Bauer, G., Brandt, M., & Kneip, T. (2022) · Journal of Happiness Studies · doi:10.1007/s10902-022-00600-8 · Cited by 21
“To address this issue, we replicated our mediation analysis in a linear SEM framework and reassessed indirect effects via health (De Stavola et al., 2015; Pearl, 2014).”
MethodGöbel, K., Niessen, C., Seufert, S., & Schmid, U. (2022) · Frontiers in Artificial Intelligence · doi:10.3389/frai.2022.919534 · Cited by 18
“…run using quasi-Bayesian Monte-Carlo simulations (we used 10,000 simulations each), which is superior to previous mediation approaches as it overcomes problems such as dependence on specific statistical models or restrictive assumptions (cf. Imai et al., 2010; Pearl, 2014; Tingley et al., 2014).”
MethodShao, Z., Wang, T., Zhang, M., Jiang, Z., Huang, S., & Zeng, P. (2021) · PLoS Computational Biology · doi:10.1371/journal.pcbi.1009250 · Cited by 15
Benchmark validation of statistical models: Application to mediation analysis of imagery and memory.
MacKinnon, D. P., Valente, M. J., & Wurpts, I. C. (2018) · Psychological Methods · doi:10.1037/met0000174 · Cited by 16
“Alternative methods to improve assessment of mediation effects have also been developed (Coffman & Zhong, 2012; Frangakis & Rubin, 2002; Imai, Keele, & Tingley, 2010; Jo, 2008; MacKinnon & Pirlott, 2015; Pearl, 2001, 2014; Valeri & VanderWeele, 2013).”
de Geus, E. J. C. (2020) · Mental health and physical activity · doi:10.1016/j.mhpa.2020.100378 · Cited by 18
Bachl, M. (2017) · Wiley eBooks · doi:10.1002/9781118901731.iecrm0038 · Cited by 18
Gender and sex bias in COVID-19 epidemiological data through the lens of causality
Díaz-Rodríguez, N., Binkytė, R., Bakkali, W., Bookseller, S., Tubaro, P., Bacevičius, A., Zhioua, S., & Chatila, R. (2023) · Information Processing & Management · doi:10.1016/j.ipm.2023.103276 · Cited by 15
“Following the interpretation of mediation analysis by Pearl (2014) we discover that being a male increases the overall risk (Total Effect Eq.”
IBCS: Learning Information Bottleneck-Constrained Denoised Causal Subgraph for Graph Classification
Yuan, R., Tang, Y., Xiao, Y., & Zhang, W. (2024) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2024.3508766 · Cited by 16
Enhanced Rock Weathering Altered Soil Organic Carbon Fluxes in a Plant Trial
Boito, L., Steinwidder, L., Rijnders, J., Berwouts, J., Janse, S., Niron, H., Roussard, J., Vienne, A., & Vicca, S. (2025) · Global Change Biology · doi:10.1111/gcb.70373 · Cited by 21
“Therefore, in order to explore the direction and magnitude of possible direct and indirect treatment effects on SCE, a causal mediation analysis by means of Monte Carlo simulations was performed on the raw SCE EGM data (Pearl 2014; Tingley et al. 2014; S8).”
Disentangled Representation for Causal Mediation Analysis
Xu, Z., Cheng, D., Li, J., Liu, J., Liu, L., & Wang, K. (2023) · Proceedings of the AAAI Conference on Artificial Intelligence · doi:10.1609/aaai.v37i9.26266 · Cited by 31
“With counterfactual thinking, defining direct and indirect effects requires no commitment to the distributional forms, and thus applies under functions with arbitrary non-linear interactions, and both continuous and categorical variables (Pearl 2014; Rubin 2005).”
BackgroundMethodInfluentialCausal relationships in longitudinal observational data: An integrative modeling approach.
Biazoli, C. E., Sato, J. R., & Pluess, M. (2024) · Psychological Methods · doi:10.1037/met0000648 · Cited by 16
Loeys, T., Talloen, W., Goubert, L., Moerkerke, B., & Vansteelandt, S. (2016) · British Journal of Mathematical and Statistical Psychology · doi:10.1111/bmsp.12077 · Cited by 20
“While assumptions (A1)–(A4) were already implicitly discussed by Judd and Kenny (1981), a number of authors used counterfactuals (Pearl, 2001; Robins & Greenland, 1992) to translate the Baron and Kenny approach within the counterfactual framework (Imai, Keele, Tingley, & Yamamoto, 2014; Muth en & Asparouhov, 2015; Pea…”
BackgroundMethodInfluentialAccommodating a Latent XM Interaction in Statistical Mediation Analysis
González, O., & Valente, M. J. (2022) · Multivariate Behavioral Research · doi:10.1080/00273171.2022.2119928 · Cited by 14
“As such, the definitions of the causal mediated effects are more general and extend to non-additive or non-linear models (Pearl, 2014).”
BackgroundPósch, K. (2019) · Sociological Methods & Research · doi:10.1177/0049124119826159 · Cited by 23
“Conversely, in nonlinear systems, the chosen level of M would influence the effect of T on Y, thus prohibiting the additivity of effects (Pearl 2014).”
BackgroundSteinberg, J. A., Somal, J., Brandel, M. G., et al. (2019) · Journal of Neurosurgical Anesthesiology · doi:10.1097/ana.0000000000000642 · Cited by 14
Group testing in mediation analysis
Derkach, A., Moore, S. C., Boca, S. M., & Sampson, J. N. (2020) · Statistics in Medicine · doi:10.1002/sim.8546 · Cited by 15
“Instead, as a pragmatic compromise, we will use the parametric models from Equations (1) and (3) and formally define s to be a mediating-set if γ jβ j ≠ 0 for at least one j∈Gs, or equivalently, if ∑ j∈Gs (βjγj)(2) ≠ 0.”
BackgroundMethodInfluentialTalloen, W., Moerkerke, B., Loeys, T., Naeghel, J. D., Keer, H. V., & Vansteelandt, S. (2016) · Journal of Educational and Behavioral Statistics · doi:10.3102/1076998616636855 · Cited by 15
Wiss, D. A., & LaFata, E. M. (2025) · Appetite · doi:10.1016/j.appet.2025.107938 · Cited by 20
“Next, continuous variables were used for all causal mediation analyses ( mediate command) using the counterfactual-based potential outcome framework (Pearl, 2014, pp. 1 – 45; VanderWeele, 2015).”
Tolera, T., & Bai, X. (2025) · Agricultural Systems · doi:10.1016/j.agsy.2025.104517 · Cited by 17
Lasch, F., Guizzaro, L., Pétavy, F., & Gallo, C. (2022) · Statistics in Biopharmaceutical Research · doi:10.1080/19466315.2022.2055633 · Cited by 14
“This allows us to apply the findings of mediation research (Pearl 2014) both to better define and to estimate the effect of interest.”
MethodChildren's Daily Routine Response to COVID-19 Emergency Measures in Serbia
Vuković, J., Matic, R. M., Milovanović, I., Maksimović, N., Krivokapić, D., & Pišot, S. (2021) · Frontiers in Pediatrics · doi:10.3389/fped.2021.656813 · Cited by 17
“Mediation analysis (20, 21) was conducted using the Sobel test (p < 0.”
MethodHuman capital in the inner city
Aliprantis, D. (2016) · Empirical Economics · doi:10.1007/s00181-016-1160-y · Cited by 16
“…“=” to denote statistical equations and identities and “←−= ” to denote structural equations.22 Given a time-invariant set of 22 I adopt Definition 5.4.1 of structural equation from Pearl (2009), so that a structural equation communicates all exclusion restrictions at a given level of measurement.”
Influential
Metadata
- Authors
- Pearl, J.
- Affiliations
- University of California, Los Angeles (US)
- Corresponding author
- Pearl, Judea
- Venue
- Psychological Methods
- Published
- 2014
- Volume
- 19
- Issue
- 4
- Pages
- 459-481
- ISSN
- 1082-989X (ISSN-L)
- 1939-1463
- OpenAlex
- W1967742189
- Cited by
- 554
- References
- 84
- Citations by year
- 2026: 45 · 2025: 52 · 2024: 49 · 2023: 49 · 2022: 56 · 2021: 57 · 2020: 87 · 2019: 38 · 2018: 32 · 2017: 33 · 2016: 25 · 2015: 21 · 2014: 10
- FWCI
- 31.34
- Type
- journal article
- Language
- en
- Open access
- green
- autopapers.ssrn.com
- Locations
- Psychological Methods (publishedVersion)
- PubMed (publishedVersion)
- SSRN Electronic Journal (submittedVersion)
- Funders
- National Institutes of Health (1R01 LM009961-01)
- National Science Foundation (IIS-0914211 and IIS-1018922)
- Office of Naval Research (N000-14-09-1-0665 and N00014-10-1-0933)
- Topics
- Physical Sciences › Mathematics › Statistics and Probability › Advanced Causal Inference Techniques
- Physical Sciences › Computer Science › Artificial Intelligence › Bayesian Modeling and Causal Inference
- Social Sciences › Arts and Humanities › History and Philosophy of Science › Philosophy of Biology and Scientific Explanation
- Keywords
- causal mediation analysis, natural direct and indirect effects, identifiability conditions, observational studies, confounding, parametric models, interaction, probabilistic graphical models
- MeSH terms
- Humans, Statistics as Topic, Models, Statistical, Causality
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.