Cited by 34,117Advanced Causal Inference TechniquesPsychometric Methodologies and TestingStatistical Methods and Bayesian Inference

Asymptotic and resampling strategies for assessing and comparing indirect effects in multiple mediator models

Open access
Preacher, Kristopher J., Hayes, Andrew F.·2008·Behavior Research Methods

Abstract

Hypotheses involving mediation are common in the behavioral sciences. Mediation exists when a predictor affects a dependent variable indirectly through at least one intervening variable, or mediator. Methods to assess mediation involving multiple simultaneous mediators have received little attention in the methodological literature despite a clear need. We provide an overview of simple and multiple mediation and explore three approaches that can be used to investigate indirect processes, as well as methods for contrasting two or more mediators within a single model. We present an illustrative example, assessing and contrasting potential mediators of the relationship between the helpfulness of socialization agents and job satisfaction. We also provide SAS and SPSS macros, as well as Mplus and LISREL syntax, to facilitate the use of these methods in applications.

AI snapshot

Written from the paper’s full text.

TL;DR

The paper argues that multiple mediation should be analyzed in one model, with the total indirect effect and each specific indirect effect tested directly rather than relying on causal steps. It recommends bootstrapping—especially bias-corrected bootstrapping—as the preferred way to obtain confidence intervals for indirect effects and contrasts.

Key findings

  • In the illustrative example with 141 employees, the total indirect effect of perceived helpfulness on job satisfaction through the three mediators was 0.1074; the asymptotic test gave var = 0.0009719, p = .0006.
  • The specific indirect effects were 0.0222 through politics, 0.0796 through people, and 0.0057 through performance; only people was statistically significant in the product-of-coefficients test (Z = 2.9366, p = .0033).
  • Bootstrapping with 5,000 resamples produced a 95% BCa CI of 0.0612 to 0.1680 for the total indirect effect; for the specific effects, only people had a CI excluding zero (0.0332 to 0.1430).
  • A pairwise contrast showed the indirect effect through people was larger than through performance; the BCa 95% CI was 0.0167 to 0.1576.

Methods

  • The paper is a methodological article on mediation in behavioral research, focused on multiple mediator models and on comparing total indirect effects, specific indirect effects, and contrasts among indirect effects.
  • It uses an illustrative example from prior research on socialization and job satisfaction, analyzed with a subsample of 141 cases and three mediators: politics, people, and performance.
  • Main analyses discussed are the causal steps approach, product-of-coefficients / delta-method tests, and bootstrapping; the paper also provides SAS, SPSS, Mplus, and LISREL syntax/macros.

Limitations

  • The paper states that the causal steps approach is of “little utility” in the multiple-mediator context because it tests individual paths rather than the indirect effect itself and yields no point estimate or SE.
  • It notes that the delta-method/product-of-coefficients approach depends on multivariate normality and that indirect effects are rarely normal in finite samples; the paper therefore treats this as a limitation of that strategy.
  • Evident from the text: the illustrative example is pedagogical and uses one subsample of 141 individuals with observed variables only, so the example is not a full latent-variable analysis and is not presented as a general empirical test of the broader theory.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.3758/brm.40.3.879

Cites

87 of 87 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Claims the paper makes

0 of 31 cited sources checked · 0 supported · 0 flagged

  • Findingown data
    § ILLUSTRATIVE EXAMPLE

    The asymptotic critical ratio for the total indirect effect leads to rejection of the null hypothesis that the total indirect effect is zero (p = .0006).

    “which leads to a rejection of the null hypothesis that the total indirect effect is zero ( p .0006).”
    Quote in the paperNot a substantial claim
  • Interpretationreasoning
    § STRATEGIES FOR ASSESSING INDIRECT EFFECTS IN MULTIPLE MEDIATOR MODELS

    Specific indirect effects can be significant even when the total indirect effect is nonsignificant.

    “It is entirely possible to find specific indirect effects to be significant in the presence of a nonsignificant total indirect effect.”
    Quote in the paperSubstantial claim
  • Interpretationnone given
    § Product-of-Coefficients Approach

    Freeing residual covariances accounts for unmodeled sources of covariation among mediators.

    “Freeing residual covariances accounts for any unmodeled sources of covariation among mediators.”
    Quote in the paperNot a substantial claim
  • Premisecitation
    § Bootstrapping

    The delta method is reasonable for very large samples, where the sampling distribution is approximately normal.

    “Sobel (1982, 1986) applied the delta method for this purpose with very large samples in mind, in which case it is reasonable to expect the sampling distribution to be approximately normal.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites Sobel, 1982, 1986

      Not checked in the free run.

  • Interpretationreasoning
    § STRATEGIES FOR ASSESSING INDIRECT EFFECTS IN MULTIPLE MEDIATOR MODELS

    The b paths in multiple mediation models are often attenuated to the degree that mediators are correlated, which can compromise the significance of specific indirect effects.

    “The effects of the mediators on Y (the b paths) are often attenuated to the degree to which the mediators are correlated, a phenomenon that can compromise the significance of particular specific indirect effects.”
    Quote in the paperSubstantial claim
  • Classified as probably not a substantial claim. 44
  • May combine two claims. 24
  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Findingown data
    § ILLUSTRATIVE EXAMPLE

    The direct effect of helpful on jobsat is 0.1057, p = .07.

    “the total and direct effects of helpful on jobsat are 0.2131, p .01, and 0.1057, p .07, respectively.”
    Quote in the paperNot a substantial claim
  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingown data
    § ILLUSTRATIVE EXAMPLE

    The specific indirect effect through politics is .0222.

    “a 1 b 1 .0222 (through politics )”
    Quote in the paperNot a substantial claim
  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingown data
    § ILLUSTRATIVE EXAMPLE

    In agreement with results of the product-of-coefficients strategy, people is the only significant mediator of the helpful → jobsat relationship.

    “In agreement with results of the product-ofcoefficients strategy, people is the only significant mediator of the helpful % jobsat relationship.”
    Quote in the paperNot a substantial claim
  • May combine two claims.

    Findingown data
    § ILLUSTRATIVE EXAMPLE

    The difference between the total and direct effects of helpful on jobsat is the total indirect effect through the three mediators, with a point estimate of .1074 and a 95% BCa bootstrap CI of 0.0612 to 0.1680.

    “The difference between the total and direct effects is the total indirect effect through the three mediators, with a point estimate of .1074 and a 95% BCa bootstrap CI of 0.0612 to 0.1680 (i.e., we can claim that the difference between the total and the direct effect of helpful on jobsat is different from zero).”
    Quote in the paperNot a substantial claim
  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingreasoning
    § ILLUSTRATIVE EXAMPLE

    The asymptotic variance of the total indirect effect is .0009719.

    “we employ Equation 1, and find that the asymptotic variance is var[ f ] .0009719.”
    Quote in the paperNot a substantial claim
  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingreasoning
    § ILLUSTRATIVE EXAMPLE

    The contrast between the indirect effects through politics and people is 0.0574.

    “Using Equation 4 to define the contrast and Equation 5 to define its sampling variance yields f c (0.2106)(0.1052) (0.2281)(0.3489) 0.0574 and var[ f c ] 0.001269.”
    Quote in the paperNot a substantial claim
  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Findingown data
    § ILLUSTRATIVE EXAMPLE

    The specific indirect effect through people is larger than the specific indirect effect through performance, with a BCa 95% CI of 0.0167 to 0.1576.

    “Examination of the pairwise contrasts of the indirect effects (C1, C2, and C3 in Appendix B) shows that the specific indirect effect through people is larger than the specific indirect effect through performance , with a BCa 95% CI of 0.0167 to 0.1576.”
    Quote in the paperNot a substantial claim
  • May combine two claims.

    Findingown data
    § ILLUSTRATIVE EXAMPLE

    The total effect of helpful on jobsat is 0.2131, p < .01.

    “the total and direct effects of helpful on jobsat are 0.2131, p .01, and 0.1057, p .07, respectively.”
    Quote in the paperNot a substantial claim
  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingown data
    § ILLUSTRATIVE EXAMPLE

    The specific indirect effect through perform is .0057.

    “a 3 b 3 .0057 (through perform )”
    Quote in the paperNot a substantial claim
  • May combine two claims.

    Findingown data
    § ILLUSTRATIVE EXAMPLE

    Only people is a mediator, since its 95% CI does not contain zero.

    “An examination of the specific indirect effects indicates that only people is a mediator, since its 95% CI does not contain zero.”
    Quote in the paperNot a substantial claim
  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingown data
    § ILLUSTRATIVE EXAMPLE

    The total indirect effect of helpful on jobsat through politics, people, and performance is .1074.

    “The total indirect effect of X on Y is f a 1 b 1 a 2 b 2 a 3 b 3 .1074.”
    Quote in the paperNot a substantial claim
  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingreasoning
    § ILLUSTRATIVE EXAMPLE

    The 95% CI for the contrast between politics and people is {-.127, .012}.

    “A 95% CI for the contrast is therefore { .0574 1.96(.001269) ½ }, or { .127, .012}.”
    Quote in the paperNot a substantial claim
  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingown data
    § ILLUSTRATIVE EXAMPLE

    The specific indirect effect through people is .0796.

    “a 2 b 2 .0796 (through people )”
    Quote in the paperNot a substantial claim
  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Findingown data
    § ILLUSTRATIVE EXAMPLE

    Neither politics nor performance contributes to the indirect effect above and beyond people.

    “Neither politics nor performance contributes to the indirect effect above and beyond people .”
    Quote in the paperNot a substantial claim
  • May combine two claims.

    Interpretationnone given
    § Product-of-Coefficients Approach

    Under maximum likelihood estimation, permitting mediator residuals to covary does not change point estimates of residual variances and coefficients, but changes the standard errors of the b paths, affecting normal-theory tests of total or specific indirect effects.

    “Under maximum likelihood estimation, the point estimates of all residual variances and coefficients will remain the same, whether or not the mediator residuals are permitted to covary. However, the SE s of the b paths will change, which in turn will affect normal-theory tests of total or specific indirect effects.”
    Quote in the paperSubstantial claim
  • May combine two claims.

    Premisecitation
    § Product-of-Coefficients Approach

    MacKinnon's (2000) standard error for the total indirect effect assumes uncorrelated mediator residuals and lacks terms containing covariances of the a paths.

    “MacKinnon (2000) provides a similar SE for the test of the total indirect effect (involving four potential mediators), but his SE assumes that the mediator residuals are uncorrelated, and consequently lacks terms containing covariances of the a paths.”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites MacKinnon, 2000

      Not checked in the free run.

  • May combine two claims.

    Premisecitation
    § MEDIATION IN BEHAVIORAL RESEARCH

    The identity c = c' + ab holds in regression and SEM when M and Y are continuous, but does not hold when one or more dependent variables are binary, requiring logistic or probit regression.

    “These identities hold in regression and structural equation modeling (SEM) where M and Y are continuous, but not in cases where one or more of the dependent variables are binary; such situations require logistic or probit regression, in which case the identity does not hold (MacKinnon & Dwyer, 1993).”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites MacKinnon & Dwyer, 1993

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

    Premisenone given
    § Product-of-Coefficients Approach

    The formulae for specific indirect effects in multiple mediation are the same as those for indirect effects in single-mediator models.

    “The formulae for specific indirect effects are the same as those for indirect effects in single-mediator models.”
    Quote in the paperNot a substantial claim
  • May combine two claims.

    Premisecitation
    § Bootstrapping

    The forced symmetry of ordinary confidence intervals results in estimation inaccuracies and problems with Type I errors and power in hypothesis testing.

    “As Efron and Tibshirani (1993) suggest, the forced symmetry of ordinary CIs results in estimation inaccuracies and problems with Type I errors and power when used in hypothesis testing.”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites Efron & Tibshirani, 1993 source

      Not checked in the free run.

  • Classified as probably not a substantial claim. May combine two claims.

    Premisecitation
    § MEDIATION IN BEHAVIORAL RESEARCH

    The distribution of the product approach and bootstrapping have higher power than the Sobel test or causal steps approach while maintaining reasonable control over the Type I error rate.

    “They recommended the use of the distribution of the product approach or bootstrapping over the Sobel test or causal steps approach, on the grounds that the former have higher power while maintaining reasonable control over the Type I error rate.”
    Quote in the paperSubstantial claim0 of 2 sources checked
    • cites MacKinnon et al., 2002 source

      Not checked in the free run.

    • cites MacKinnon et al., 2004 source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S4 (based).

    Premisenone given
    § Distribution of the Product Strategy

    The distribution of sums or differences of products, necessary for testing total indirect effects and pairwise contrasts among indirect effects, has not yet been worked out.

    “Whereas the distribution of the product strategy can be used for testing specific indirect effects, the distribution of sums or differences of products (necessary for testing hypotheses about total indirect effects and pairwise contrasts among indirect effects, to be discussed later) has not yet been worked out.”
    Quote in the paperNot a substantial claim
  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisenone given
    § Bootstrapping

    There are only two empirical investigations of the performance of bootstrapping in a multiple-mediator context.

    “To our knowledge, there are only two empirical investigations of the performance of bootstrapping in a multiplemediator context.”
    Quote in the paperNot a substantial claim
  • May combine two claims.

    Interpretationnone given
    § Product-of-Coefficients Approach

    If the parameters are free, the multiple mediation model is saturated (df = 0) and the a path coefficients will typically have nonzero covariances.

    “If these parameters are free, the model is saturated (i.e., df 0) and the a path coefficients will typically have nonzero covariances.”
    Quote in the paperNot a substantial claim
  • May combine two claims.

    Premisecitation
    § Bootstrapping

    Bootstrapping generally was superior to the multivariate product-of-coefficients strategy in small to moderate samples, with BC and BCa performing best in terms of both power and Type I error rates.

    “She found that bootstrapping generally was superior to the multivariate product-of-coefficients strategy in small to moderate samples, with BC and BCa performing best in terms of both power and Type I error rates.”
    Quote in the paperSubstantial claim
  • May combine two claims.

    Interpretationnone given
    § Bootstrapping

    The sampling distribution of the bootstrap product ai bi* is skewed relative to a normal distribution unless a = b = 0, so confidence limits should not be equidistant from the point estimate.

    “The sampling distribution of ai bi * is skewed relative to a normal distribution (unless a b 0), and hence the confidence limits should not be equidistant from the point estimate.”
    Quote in the paperSubstantial claim
  • May combine two claims.

    Interpretationreasoning
    § Causal Steps Approach

    The causal steps approach in multiple mediation relies on tests of individual a and b paths rather than testing specific indirect effects, and yields no point estimate or standard error of the mediation effect.

    “It relies on a set of tests of individual a and b paths rather than testing the specific indirect effects, and yields no point estimate or SE of the mediation effect.”
    Quote in the paperSubstantial claim
  • Classified as probably not a substantial claim.

    Premisecitation
    § Bootstrapping

    Bootstrapping generally was superior to the multivariate product-of-coefficients strategy in a two-mediator model.

    “She found that bootstrapping generally was superior to the multivariate product-of-coefficients”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites Briggs, 2006

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Interpretationnone given
    § SOFTWARE IMPLEMENTATION

    At the user's request, the macro conducts all possible pairwise contrasts of indirect effects in multiple mediator models.

    “Additionally, at the user' s request, the macro conducts all possible pairwise contrasts of indirect effects in multiple mediator models.”
    Quote in the paperNot a substantial claim
  • Probably not a substantial claim: may not pass S2 (checkable), S4 (based).

    Interpretationnone given
    § Recommendations

    In most situations, it is unlikely that the effect of an independent variable on an outcome is transmitted by only one means.

    “In most situations, it is unlikely that the effect of an independent variable on an outcome is transmitted by only one means.”
    Quote in the paperNot a substantial claim
  • Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing).

    Interpretationreasoning
    § NOTES

    Such paradoxes occur frequently.

    “Such paradoxes occur frequently and are a result, in part, of thinking about effects in dichotomous terms-as zero or something else.”
    Quote in the paperNot a substantial claim
  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Interpretationnone given
    § Extensions

    The macros can be used to estimate indirect effects in models with k independent variables by estimating k models, with one variable chosen as the independent variable and the remaining k-1 variables treated as covariates.

    “The macros can also be used to estimate indirect effects in models with k independent variables by estimating k models. In each of these models, one of the variables is chosen as the independent variable, and the remaining k 1 variables are treated as covariates.”
    Quote in the paperNot a substantial claim
  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Premisenone given
    § Extensions

    SEM explicitly models measurement error, allowing researchers to test hypotheses using latent constructs rather than imperfect measured indicators.

    “One of the most attractive features of SEM is that, unlike regression analysis of the type performed with SAS or SPSS, it explicitly models measurement error, allowing researchers to test hypotheses using the latent constructs rather than imperfect measured indicators.”
    Quote in the paperNot a substantial claim
  • Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

    Premisenone given
    § Extensions

    Many SEM packages now include delta method estimates of SEs for indirect effects.

    “Many SEM packages now include delta method estimates of SE s for indirect effects.”
    Quote in the paperNot a substantial claim
  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Interpretationown data
    § ILLUSTRATIVE EXAMPLE

    The directions of the a and b paths are consistent with the interpretation that greater helpfulness leads to greater mastery of the politics, people, and performance dimensions, which in turn leads to greater job satisfaction.

    “From output in Appendix B, the directions of the a and b paths are consistent with the interpretation that greater helpfulness leads to greater mastery of the politics , people , and performance dimensions, which in turn leads to greater job satisfaction.”
    Quote in the paperNot a substantial claim
  • May combine two claims.

    Premisecitation
    § CONTRASTING INDIRECT EFFECTS IN MULTIPLE MEDIATOR MODELS

    Constraining residual covariances to zero results in independent a paths, yielding the formula derived by MacKinnon (2000) for the variance of a single pairwise contrast.

    “Constraining residual covariances to zero results in independent a paths, yielding the formula derived by MacKinnon (2000) for the variance of a single pairwise contrast,”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites MacKinnon (2000)

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S4 (based).

    Interpretationnone given
    § Recommendations

    The product-of-coefficients strategy is useful only when the assumption of normality of the sampling distribution of the indirect effect can be reasonably met, as when large samples are available or when the effects are large.

    “The product-of-coefficients strategy is useful only when the assumption of normality of the sampling distribution of the indirect effect can be reasonably met, as when large samples are available or when the effects are large.”
    Quote in the paperNot a substantial claim
  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Interpretationreasoning
    § NOTES

    It is important to minimize sampling variance in estimates of percentile endpoints and acceleration constant.

    “Given that the percentile endpoints are estimated using extremes of the sampling distribution and that, for BCa intervals, a good estimate of the acceleration constant is needed, it is important to minimize sampling variance in these estimates.”
    Quote in the paperNot a substantial claim
    • Rests on an assumption: Percentile endpoints are estimated using extremes of the sampling distribution.
    • Rests on an assumption: For BCa intervals, a good estimate of the acceleration constant is needed.
  • Classified as probably not a substantial claim.

    Interpretationreasoning
    § Recommendations

    Contrasts do not compare two mediators in their ability to mediate, but rather their unique abilities to mediate, above and beyond any other mediators or covariates in the model.

    “Another consequence of the fact that mediators are usually correlated is that contrasts do not compare two mediators in their ability to mediate, but rather their unique abilities to mediate, above and beyond any other mediators or covariates in the model.”
    Quote in the paperSubstantial claim
  • Probably not a substantial claim: may not pass S2 (checkable), S4 (based).

    Interpretationnone given
    § Recommendations

    When multiple mediators are entertained, it is often more convenient, precise, and parsimonious to include all of them in the same model.

    “When multiple mediators are entertained, it is often more convenient, precise, and parsimonious to include all of them in the same model.”
    Quote in the paperNot a substantial claim
  • Classified as probably not a substantial claim. May combine two claims.

    Interpretationnone given
    § Recommendations

    A specific indirect effect does not represent the ability of a given mediator M to mediate the effect of X on Y; rather, it represents the ability of M to mediate the effect controlling for all other mediators.

    “A specific indirect effect does not represent the ability of a given mediator M to mediate the effect of X on Y . Rather, a specific indirect effect represents the ability of M to mediate the effect controlling for all other mediators.”
    Quote in the paperSubstantial claim
  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Interpretationreasoning
    § NOTES

    Such paradoxes are a result, in part, of thinking about effects in dichotomous terms-as zero or something else.

    “Such paradoxes occur frequently and are a result, in part, of thinking about effects in dichotomous terms-as zero or something else.”
    Quote in the paperNot a substantial claim
  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Interpretationreasoning
    § NOTES

    Such paradoxes result from conducting multiple tests that differ in power.

    “Ultimately, such paradoxes result from conducting multiple tests that differ in power.”
    Quote in the paperNot a substantial claim
  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Interpretationnone given
    § SOFTWARE IMPLEMENTATION

    The macro allows for statistical control of one or more covariates that are not proposed to be mediators of the total effect.

    “It also allows for statistical control of one or more covariates that are not proposed to be mediators of the total effect.”
    Quote in the paperNot a substantial claim
  • Classified as probably not a substantial claim.

    Interpretationcitation
    § Recommendations

    Bootstrapping provides the most powerful and reasonable method of obtaining confidence limits for specific indirect effects under most conditions.

    “Bootstrapping provides the most powerful and reasonable method of obtaining confidence limits for specific indirect effects under most conditions, so our primary recommendation is to use bootstrapping-in particular, BC bootstrapping-whenever possible; see Briggs (2006), Williams (2004), and Williams and MacKinnon (2008) for extensive simulation results supporting this recommendation.”
    Quote in the paperSubstantial claim0 of 3 sources checked
    • cites Briggs (2006)

      Not checked in the free run.

    • cites Williams (2004)

      Not checked in the free run.

    • cites Williams and MacKinnon (2008) source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S2 (checkable).

    Interpretationreasoning
    § Recommendations

    A specific indirect effect for M2 represents M2's unique ability to mediate the X→Y relationship.

    “Thus, a specific indirect effect for M 2 represents M 2 ' s unique ability to mediate the X % Y relationship.”
    Quote in the paperNot a substantial claim
  • Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing).

    Interpretationreasoning
    § NOTES

    More bootstrap samples is better.

    “There is no consensus as to how many bootstrap samples should be generated, except that more is better.”
    Quote in the paperNot a substantial claim
  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation
    § SOFTWARE IMPLEMENTATION

    Some programs, such as LISREL and Mplus, can be programmed to conduct contrasts of indirect effects.

    “Some programs, such as LISREL and Mplus, can be programmed to conduct contrasts of indirect effects (Cheung, 2007).”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites Cheung, 2007

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation
    § SOFTWARE IMPLEMENTATION

    AMOS implements the percentile bootstrap method for total indirect effects in simple and multiple mediator models.

    “AMOS (Arbuckle, 1999) implements the percentile bootstrap method for total indirect effects in simple and multiple mediator models.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites Arbuckle, 1999

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Interpretationnone given
    § SOFTWARE IMPLEMENTATION

    The user can request tests of total and specific indirect effects using the product-of-coefficients approach or by bootstrapping confidence intervals (percentile, BC, and BCa) at any desired confidence level.

    “The user can request tests of total and specific indirect effects using the product-of-coefficients approach or by bootstrapping confidence intervals (percentile, BC, and BCa) at any desired confidence level (Mplus is inflexible in this regard).”
    Quote in the paperNot a substantial claim
  • Probably not a substantial claim: may not pass S2 (checkable).

    Interpretationreasoning
    § Extensions

    Tests of specific indirect effects, whether by using the product-of-coefficients approach or the resampling approach, would facilitate teasing apart the separate roles played by individual intervening variables.

    “Under such circumstances, tests of specific indirect effects, whether by using the product-of-coefficients approach or the resampling approach, would facilitate teasing apart the separate roles played by individual intervening variables.”
    Quote in the paperNot a substantial claim
  • Probably not a substantial claim: may not pass S4 (based). May combine two claims.

    Interpretationnone given
    § Recommendations

    Collinearity may lead the investigator to conclude that M does not serve as a mediator when in fact it does, or even to conclude that M serves as a mediator when it does not.

    “Collinearity is not necessarily a problem, but it may lead the investigator to conclude that M does not serve as a mediator when in fact it does, or even to conclude that M serves as a mediator when it does not.”
    Quote in the paperNot a substantial claim
  • Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing).

    Interpretationreasoning
    § Recommendations

    Following the strategy of selecting mediators that represent unique constructs with as little conceptual overlap as possible will minimize collinearity.

    “Following this strategy will minimize collinearity.”
    Quote in the paperNot a substantial claim
  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Premisecitation
    § CONTRASTING INDIRECT EFFECTS IN MULTIPLE MEDIATOR MODELS

    MacKinnon (2000) discusses several other useful contrasts and derives SEs for them under the assumption that the mediator residual covariances are constrained to zero.

    “MacKinnon (2000) discusses several other useful contrasts and derives SE s for them under the assumption that the mediator residual covariances are constrained to zero.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites MacKinnon (2000)

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Interpretationreasoning
    § SOFTWARE IMPLEMENTATION

    The authors have developed macros for SPSS and SAS that provide researchers who prefer these programs the ability to conduct the type of analysis described in the paper.

    “We have developed macros for SPSS and SAS that provide researchers who prefer these programs the ability to conduct the type of analysis we have described here.”
    Quote in the paperNot a substantial claim
  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Interpretationnone given
    § Extensions

    The authors' SPSS and SAS macros can be used for multiple mediation in the SEM special case of path analysis, in which all variables are observed.

    “Our SPSS and SAS macros can be used for multiple mediation in the SEM special case of path analysis, in which all variables are observed.”
    Quote in the paperNot a substantial claim
  • Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

    Interpretationnone given
    § Extensions

    The results generalize to SEM with latent variables.

    “It should be emphasized that the results generalize to SEM with latent variables.”
    Quote in the paperNot a substantial claim
  • May combine two claims.

    Interpretationnone given
    § CONTRASTING INDIRECT EFFECTS IN MULTIPLE MEDIATOR MODELS

    The square root of the variance of a contrast is the first-order SE of the contrast, assuming normality for the contrast, and can be used to test hypotheses or construct CIs.

    “As with specific and total indirect effects, the square root of this variance is the first-order SE of the contrast, assuming normality for the contrast, and can thus be used to test hypotheses or to construct CIs.”
    Quote in the paperSubstantial claim
  • May combine two claims.

    Premisecitation
    § Bootstrapping

    There were differences between the percentile and bias corrected methods, depending on the size of the paths in the model, with the percentile method generally showing a slight superiority in conditions in which the bias corrected method was slightly liberal.

    “There were differences between the percentile and bias corrected methods, depending on the size of the paths in the model, with the percentile method generally showing a slight superiority in conditions in which the bias corrected method was slightly liberal.”
    Quote in the paperNot a substantial claim0 of 2 sources checked
    • cites Williams (2004)

      Not checked in the free run.

    • cites Williams and MacKinnon (2008) source

      Not checked in the free run.

58 of 58
  • The paper relies on this without support.

    The causal order of X, M, and Y can be established on theoretical or procedural grounds.

    “We address only the situation in which the causal order of X , M , and Y can be established on theoretical or procedural grounds.”
  • The paper relies on this without support.

    The sampling distribution of ab is normal only in large samples.

    “Methodologists have taken issue with the use of the standard normal distribution for deriving a p value for the indirect effect, since the sampling distribution of ab is normal only in large samples.”
  • The paper relies on this without support.

    Point estimates of path coefficients do not depend on the choice to free or constrain residual covariances.

    “point estimates of path coefficients do not depend on the choice to free or constrain residual covariances”
  • The paper relies on this without support.

    Many researchers still rely on historically popular programs, such as SPSS or SAS, for their analyses.

    “many researchers still rely on historically popular programs, such as SPSS or SAS, for their analyses.”
  • The paper relies on this without support.

    Neither SPSS nor SAS has the intrinsic capability to conduct the type of analysis described in the paper.

    “Neither SPSS nor SAS has the intrinsic capability to conduct the type of analysis we have described here”
  • The paper relies on this without support.

    There is no consensus as to how many bootstrap samples should be generated.

    “There is no consensus as to how many bootstrap samples should be generated, except that more is better.”
  • The paper relies on this without support.

    1,000 bootstrap resamples is probably sufficient for preliminary analyses.

    “although 1,000 is probably sufficient for preliminary analyses”
  • The paper relies on this without support.

    A pattern may occur in models containing both a mediation effect and a suppression effect where either effect is large and significant but their sum is small and nonsignificant.

    “This pattern may occur in models containing both a mediation effect and a suppression effect. Even though either effect may be large and significant, their sum may be small and nonsignificant.”
  • The paper relies on this without support.

    Percentile endpoints are estimated using extremes of the sampling distribution.

    “Given that the percentile endpoints are estimated using extremes of the sampling distribution”
  • The paper relies on this without support.

    For BCa intervals, a good estimate of the acceleration constant is needed.

    “and that, for BCa intervals, a good estimate of the acceleration constant is needed”
10 of 10

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

  • Simple mediation analyses accompanied the multiple mediation analysis to bypass suppression effects.

    “As has been recommended (see MacKinnon, Krull, & Lockwood, 2000; Preacher & Hayes, 2008), the multiple mediation analysis was accompanied by simple mediation analyses to bypass the issues that suppression effects can obscure possible indirect effects of individual mediators.”

    From Change in Decentering Mediates Improvement in Anxiety in Mindfulness-Based Stress Reduction for Generalized Anxiety Disorder (2015) · cites this paper as Preacher & Hayes, 2008

    The Recommendations section says correlated mediators can obscure whether an individual variable mediates an effect. But the paper recommends testing specific indirect effects within a multiple-mediator model, not accompanying that model with separate simple mediation analyses; the Strategies section warns that separate simple models can produce biased estimates. (LLM, conf=0.87)

    Experimental — accuracy not yet measured
  • Bootstrapping with 5,000 iterations showed the mediating effect of self-esteem was significant, 95% CI [-.1807, -.0215], excluding 0.

    “The 95% confidence interval was [-.1807, -.0215], excluding 0, which indicates that the mediating effect of self-esteem was significant.”

    From Social media addiction: Its impact, mediation, and intervention (2019) · cites this paper as Preacher & Hayes, 2008

    The Bootstrapping section explains how to construct 95% confidence intervals for indirect effects, and the Illustrative Example treats an effect as significant when its interval excludes zero. The example uses 5,000 bootstrap samples, but it concerns other mediators; our copy does not report a self-esteem effect or the interval [-.1807, -.0215]. (LLM, conf=0.88)

    Experimental — accuracy not yet measured
  • Indirect (mediation) effects were assessed using the product-of-coefficients approach with robust standard errors, appropriate under DWLS estimation.

    “Indirect (mediation) effects were assessed using the product-of-coefficients approach with robust standard errors [54], [55], appropriate under DWLS estimation.”

    From Enhancing Digital Financial Literacy and Digital Trust for Sustainable Online Loan Adoption among Micro-Entrepreneurs in Developing Economies (2026) · cites this paper as [54]

    The Product-of-Coefficients Approach section describes assessing indirect effects using coefficient products and delta-method standard errors. It does not establish the use of robust standard errors or the approach’s appropriateness under DWLS estimation; its discussion of estimation specifically mentions maximum likelihood. (LLM, conf=0.88)

    Experimental — accuracy not yet measured
  • Bootstrap mediation analyses can be applied when the sample size is small or moderate (e.g., n = 20-60).

    “Bootstrap mediation analyses can be applied when the sample size is small or moderate (e.g., n = 20–60) [79, 80].”

    From Improving symptom management for survivors of young adult cancer: rationale and study protocol for a pilot randomized controlled trial (2024) · cites this paper as [79]

    The Bootstrapping section reports that bootstrapping generally outperformed the product-of-coefficients method in small to moderate samples, and the Recommendations section advocates its use whenever possible. Our copy does not specify that this applies to samples of n = 20–60, so the numerical example is not verified. (LLM, conf=0.85)

    Experimental — accuracy not yet measured
  • Because unstandardized beta coefficients were used, differences in scale magnitude among variables do not affect the estimation of indirect effects [47].

    “Because unstandardized beta coefficients were used, differences in scale magnitude among variables do not affect the estimation of indirect effects [47].”

    From Four Facet Mindful Eating and Eating Disorder Risk: Mediation By Depression and Body Mass Index. (2026) · cites this paper as [47]

    The ‘Contrasting Indirect Effects in Multiple Mediator Models’ section says that multiplying the unstandardized path coefficients cancels the mediator’s scale, allowing indirect effects through different mediators to be compared for the same independent–dependent variable pair. It does not say that scale differences among all variables leave indirect-effect estimates unaffected; the claim omits the restriction to mediator scales and a shared independent–dependent variable pair. Page information is unavailable in our copy. (LLM, conf=0.88)

    Experimental — accuracy not yet measured

Cites

Flagged first, then by citations

Research works printed but not linked: 5

  1. Arbuckle, J. L. (1999). AMOS 4 [Computer software]. Chicago: Small-Waters Corp.
  2. Lockwood, C. M., & MacKinnon, D. P. (1998). Bootstrapping the standard error of the mediated effect. Proceedings of the 23rd Annual Meeting of SAS Users Group International (pp. 997–1002). Cary, NC: SAS Institute, Inc.
  3. Muthén, L. K., & Muthén, B. O. (2007). Mplus user’s guide (5th ed.). Los Angeles: Muthén & Muthén.
  4. Neale, M. C., Boker, S. M., Xie, G., & Maes, H. H. (2003). Mx: Statistical modeling (6th ed.). Richmond, VA: Department of Psychiatry, Virginia Commonwealth University.
  5. Williams, J. (2004). Resampling and distribution of the product methods for testing indirect effects in complex models. Unpublished doctoral dissertation, Arizona State University, Tempe, AZ.

Cited by

Sorted by citations

Authors
Preacher, K. J., & Hayes, A. F.
Affiliations
  • University of Kansas (US)
  • The Ohio State University (US)
Corresponding author
Preacher, Kristopher J.
Venue
Behavior Research Methods
Published
2008-08
Volume
40
Issue
3
Pages
879-891
ISSN
  • 1554-351X (ISSN-L)
  • 1554-3528
OpenAlex
W2038702827
Cited by
34,117
References
84
Citations by year
2026: 3186 · 2025: 2505 · 2024: 2176 · 2023: 2211 · 2022: 2513 · 2021: 2399 · 2020: 2173 · 2019: 2057 · 2018: 2036 · 2017: 2079 · 2016: 2060 · 2015: 2059 · 2014: 1902 · 2013: 1687 · 2012: 1326
FWCI
533.48
Type
journal article
Language
en
Open access
Locations
  • Behavior Research Methods (publishedVersion)
  • PubMed (publishedVersion)
Funders
  • National Institutes of Health (5f32da016883-03)
  • National Institute on Drug Abuse (DA16883)
Topics
  • Physical Sciences › Mathematics › Statistics and Probability › Advanced Causal Inference Techniques
  • Social Sciences › Decision Sciences › Management Science and Operations Research › Psychometric Methodologies and Testing
  • Physical Sciences › Mathematics › Statistics and Probability › Statistical Methods and Bayesian Inference
Keywords
multiple mediation, indirect effects, multiple mediators, asymptotic methods, resampling, bootstrapping, job satisfaction, socialization agents
MeSH terms
Behavioral Medicine, Humans, Social Behavior, Socialization, Software, Models, Statistical

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.

Claims read by deepseek-v4.1-flash under definition version 3 (extract-v3) on 2026-10-11.

Classified by jev-1.13-20260917 (gold-guideline-v0.2) on 2026-10-11.

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