Questions of value, questions of magnitude: An exploration and application of methods for comparing indirect effects in multiple mediator models.

Mediation analysis is widely used to test and inform theory and debate about the mechanism(s) by which causal effects operate, quantitatively operationalized as an indirect effect in a mediation model. Most effects operate through multiple mechanisms simultaneously, and a mediation model is likely to be more realistic when it is specified to capture multiple mechanisms at the same time with the inclusion of more than one mediator in the model. This also allows an investigator to compare indirect effects to each other. After an overview of the mechanics of mediation analysis, we advocate formally comparing indirect effects in models that include more than one mediator, focusing on the important distinction between questions and claims about value (i.e., are two indirect effects the same number?) versus magnitude (i.e., are two indirect effects equidistant from zero or the same in strength?). After discussing the shortcomings of the conventional method for comparing two indirect effects in a multiple mediator model-which only answers a question about magnitude in some circumstances-we introduce several methods that, unlike the conventional approach, always answer questions about difference in magnitude. We illustrate the use of these methods and provide code that implements them in popular software. We end by summarizing simulation findings and recommending which method(s) to prefer when comparing like- and opposite-signed indirect effects.

Mediation, Moderation, and Conditional Process Analysis: Concepts, Computations, and Some Common Confusions.

This work provides a conceptual introduction to mediation, moderation, and conditional process analysis in psychological research. We discuss the concepts of direct effect, indirect effect, total effect, conditional effect, conditional direct effect, conditional indirect effect, and the index of moderated mediation index, while providing our perspective on certain analysis and interpretation confusions that sometimes arise in practice in this journal and elsewhere, such as reliance on the causal steps approach and the Sobel test in mediation analysis, misinterpreting the regression coefficients in a model that includes a product of variables, and subgroups mediation analysis rather than conditional process analysis when exploring whether an indirect effect depends on a moderator. We also illustrate how to conduct various analyses that are the focus of this paper with the freely-available PROCESS procedure available for SPSS, SAS, and R, using data from an experimental investigation on the effectiveness of personal or testimonial narrative messages in improving intergroup attitudes.

Two-condition within-participant statistical mediation analysis: A path-analytic framework.

Researchers interested in testing mediation often use designs where participants are measured on a dependent variable Y and a mediator M in both of 2 different circumstances. The dominant approach to assessing mediation in such a design, proposed by Judd, Kenny, and McClelland (2001), relies on a series of hypothesis tests about components of the mediation model and is not based on an estimate of or formal inference about the indirect effect. In this article we recast Judd et al.'s approach in the path-analytic framework that is now commonly used in between-participant mediation analysis. By so doing, it is apparent how to estimate the indirect effect of a within-participant manipulation on some outcome through a mediator as the product of paths of influence. This path-analytic approach eliminates the need for discrete hypothesis tests about components of the model to support a claim of mediation, as Judd et al.'s method requires, because it relies only on an inference about the product of paths-the indirect effect. We generalize methods of inference for the indirect effect widely used in between-participant designs to this within-participant version of mediation analysis, including bootstrap confidence intervals and Monte Carlo confidence intervals. Using this path-analytic approach, we extend the method to models with multiple mediators operating in parallel and serially and discuss the comparison of indirect effects in these more complex models. We offer macros and code for SPSS, SAS, and Mplus that conduct these analyses. (PsycINFO Database Record

Regression-based statistical mediation and moderation analysis in clinical research: Observations, recommendations, and implementation.

There have been numerous treatments in the clinical research literature about various design, analysis, and interpretation considerations when testing hypotheses about mechanisms and contingencies of effects, popularly known as mediation and moderation analysis. In this paper we address the practice of mediation and moderation analysis using linear regression in the pages of Behaviour Research and Therapy and offer some observations and recommendations, debunk some popular myths, describe some new advances, and provide an example of mediation, moderation, and their integration as conditional process analysis using the PROCESS macro for SPSS and SAS. Our goal is to nudge clinical researchers away from historically significant but increasingly old school approaches toward modifications, revisions, and extensions that characterize more modern thinking about the analysis of the mechanisms and contingencies of effects.

An Index and Test of Linear Moderated Mediation.

I describe a test of linear moderated mediation in path analysis based on an interval estimate of the parameter of a function linking the indirect effect to values of a moderator-a parameter that I call the index of moderated mediation. This test can be used for models that integrate moderation and mediation in which the relationship between the indirect effect and the moderator is estimated as linear, including many of the models described by Edwards and Lambert ( 2007 ) and Preacher, Rucker, and Hayes ( 2007 ) as well as extensions of these models to processes involving multiple mediators operating in parallel or in serial. Generalization of the method to latent variable models is straightforward. Three empirical examples describe the computation of the index and the test, and its implementation is illustrated using Mplus and the PROCESS macro for SPSS and SAS.

Injury news coverage, relative concern, and support for alcohol-control policies: an impersonal impact explanation.

Research on the impersonal impact hypothesis suggests that news (especially print) coverage of health and safety risks primarily influences perceptions of risk as a societal issue, and not perceptions of personal risk. The authors propose that the impersonal impact of news-impact primarily on concerns about social-level risks-will mediate effects of news stories on support for public health policies; such effects substantively matter as evidence suggests health policies, in turn, have important effects on protective behaviors and health outcomes. In an experiment using 60 randomly selected violent crime and accident news stories manipulated to contain or not contain reference to alcohol use as a causative factor, the authors find that the effect of stories that mention alcohol as a causative factor on support for alcohol-control policies is mediated by social-level concern and not by personal-level concern. In so doing, the authors provide a theoretical explanation as well as empirical evidence regarding the potential for news coverage-including breaking or episodic news-to influence health-related public policy.

Statistical mediation analysis with a multicategorical independent variable.

Virtually all discussions and applications of statistical mediation analysis have been based on the condition that the independent variable is dichotomous or continuous, even though investigators frequently are interested in testing mediation hypotheses involving a multicategorical independent variable (such as two or more experimental conditions relative to a control group). We provide a tutorial illustrating an approach to estimation of and inference about direct, indirect, and total effects in statistical mediation analysis with a multicategorical independent variable. The approach is mathematically equivalent to analysis of (co)variance and reproduces the observed and adjusted group means while also generating effects having simple interpretations. Supplementary material available online includes extensions to this approach and Mplus, SPSS, and SAS code that implements it.

The relative trustworthiness of inferential tests of the indirect effect in statistical mediation analysis: does method really matter?

A content analysis of 2 years of Psychological Science articles reveals inconsistencies in how researchers make inferences about indirect effects when conducting a statistical mediation analysis. In this study, we examined the frequency with which popularly used tests disagree, whether the method an investigator uses makes a difference in the conclusion he or she will reach, and whether there is a most trustworthy test that can be recommended to balance practical and performance considerations. We found that tests agree much more frequently than they disagree, but disagreements are more common when an indirect effect exists than when it does not. We recommend the bias-corrected bootstrap confidence interval as the most trustworthy test if power is of utmost concern, although it can be slightly liberal in some circumstances. Investigators concerned about Type I errors should choose the Monte Carlo confidence interval or the distribution-of-the-product approach, which rarely disagree. The percentile bootstrap confidence interval is a good compromise test.

Self-efficacy mediates the impact of craving on smoking abstinence in low to moderately anxious patients: results of a moderated mediation approach.

Little is known about the effect of craving on smoking abstinence among cardiac patients who smoked prior to admission and the mechanisms that might facilitate success in smoking cessation after discharge from hospital. This study examined the mediating effect of self-efficacy on the relationship between craving and smoking abstinence and how this mechanism may be contingent on emotional state at the time of hospital admission. Cardiac patients who smoked prior to admission were recruited from cardiac nursing units in Dutch hospitals. On hospitalization, 244 patients completed a questionnaire on craving, self-efficacy to smoking cessation, and anxiety and depression levels. Six months after discharge patients were interviewed to ascertain their smoking status. Simple mediation and moderated mediation effects of craving and self-efficacy on smoking abstinence were tested. Of the patients who successfully completed the baseline questionnaire and the follow-up interview, 38% were not smoking at 6 months. Self-efficacy mediated the effect of craving on smoking abstinence. However, this indirect effect was more pronounced among patients with relatively low to moderate anxiety at the time of hospitalization. Our findings suggest that craving reduces self-efficacy, which in turn reduces the likelihood of smoking abstinence, although this process applies only to those patients with low to moderate anxiety levels at the time of hospitalization. Interventions for smoking cardiac patients should aim to reduce craving and to enhance patients' self-efficacy to smoking cessation after discharge from hospital.

Increasing support for alcohol-control enforcement through news coverage of alcohol's role in injuries and crime.

OBJECTIVE: Prior research has shown that the proportion of news stories about violent crimes, car crashes, and other unintended injuries that mention the possible contributing role of alcohol is far lower than the actual proportion of alcohol-related crimes and unintended injuries. An experiment was conducted to test the hypothesis that inclusion of such mention can increase concern about alcohol risks and support for alcohol-control measures, which have elsewhere been shown to decrease alcohol-related problems in community settings. Methodologically, we provide a model for experiments permitting generalization across randomly selected message stimuli. METHOD: Sixty randomly selected local news stories on violent crime, motor vehicle crashes, and other unintended injuries from newspapers throughout the United States were manipulated into versions including or not including alcohol as a causative factor. Participants (n = 785) were drawn from a national online research panel representative of the U.S. population; 66% of panel members contacted agreed to participate. Data were analyzed using mixed-effect, multilevel models to permit generalization across message and participant variability. RESULTS: Mention of alcohol in news stories increased support for enforcement of alcohol-control laws. CONCLUSIONS: Efforts to increase mention of alcohol as a causative factor in news reports of violent crime and unintended injury have the potential to increase public support for alcohol-control policies.

The Influence of Youth Music Television Viewership on Changes in Cigarette Use and Association with Smoking Peers: A Social Identity, Reinforcing Spirals Perspective.

Prior research has found strong evidence of a prospective association between R movie exposure and teen smoking. Using parallel process latent-growth modeling, the present study examines prospective associations between viewing of music video channels on television (e.g., MTV and VH-1) and changes over time in smoking and association with smoking peers. Results showed that baseline viewing of music-oriented channels such as MTV and VH-1 robustly predicted increasing trajectories of smoking and of associating with smoking peers, even after application of a variety of controls including parent reports of monitoring behavior. These results are consistent with the arguments from the reinforcing spirals model that such media use serves as a means of developing emergent adolescent social identities consistent with associating with smoking peers and acquiring smoking and other risk behaviors; evidence also suggests that media choice in reinforcing spiral processes are dynamic and evolve as social identity evolves.

Quantifying and Testing Indirect Effects in Simple Mediation Models When the Constituent Paths Are Nonlinear.

Most treatments of indirect effects and mediation in the statistical methods literature and the corresponding methods used by behavioral scientists have assumed linear relationships between variables in the causal system. Here we describe and extend a method first introduced by Stolzenberg (1980) for estimating indirect effects in models of mediators and outcomes that are nonlinear functions but linear in their parameters. We introduce the concept of the instantaneous indirect effect of X on Y through M and illustrate its computation and describe a bootstrapping procedure for inference. Mplus code as well as SPSS and SAS macros are provided to facilitate the adoption of this approach and ease the computational burden on the researcher.

Computational procedures for probing interactions in OLS and logistic regression: SPSS and SAS implementations.

Researchers often hypothesize moderated effects, in which the effect of an independent variable on an outcome variable depends on the value of a moderator variable. Such an effect reveals itself statistically as an interaction between the independent and moderator variables in a model of the outcome variable. When an interaction is found, it is important to probe the interaction, for theories and hypotheses often predict not just interaction but a specific pattern of effects of the focal independent variable as a function of the moderator. This article describes the familiar pick-a-point approach and the much less familiar Johnson-Neyman technique for probing interactions in linear models and introduces macros for SPSS and SAS to simplify the computations and facilitate the probing of interactions in ordinary least squares and logistic regression. A script version of the SPSS macro is also available for users who prefer a point-and-click user interface rather than command syntax.

Newspaper Coverage of Cancer Prevention: Multilevel Evidence for Knowledge Gap Effects.

Prior research on knowledge gap effects, in health as well as in other domains, has focused largely on assessing individual-level differences in exposure to news based on self-report of media use. Inherent inferential limitations of this approach are addressed by testing the hypothesis that the relationship between education and cancer prevention knowledge will be moderated by regional differences in U.S. news coverage of cancer prevention. The study also tests, using these methods, findings by Kwak (1999) suggesting that the importance of attention to relevant news in predicting knowledge decreases as information available in the news increases. Using a representative national sample of newspaper coverage to assess regional differences in cancer prevention coverage, a representative national probability sample to assess respondent education and cancer prevention knowledge, and multilevel analyses of the relationship between regional coverage differences and knowledge of persons in those regions, support is found for both of these propositions.

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

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.

Further evaluating the conditional decision rule for comparing two independent means.

Many books on statistical methods advocate a 'conditional decision rule' when comparing two independent group means. This rule states that the decision as to whether to use a 'pooled variance' test that assumes equality of variance or a 'separate variance' Welch t test that does not should be based on the outcome of a variance equality test. In this paper, we empirically examine the Type I error rate of the conditional decision rule using four variance equality tests and compare this error rate to the unconditional use of either of the t tests (i.e. irrespective of the outcome of a variance homogeneity test) as well as several resampling-based alternatives when sampling from 49 distributions varying in skewness and kurtosis. Several unconditional tests including the separate variance test performed as well as or better than the conditional decision rule across situations. These results extend and generalize the findings of previous researchers who have argued that the conditional decision rule should be abandoned.

Using heteroskedasticity-consistent standard error estimators in OLS regression: an introduction and software implementation.

Homoskedasticity is an important assumption in ordinary least squares (OLS) regression. Although the estimator of the regression parameters in OLS regression is unbiased when the homoskedasticity assumption is violated, the estimator of the covariance matrix of the parameter estimates can be biased and inconsistent under heteroskedasticity, which can produce significance tests and confidence intervals that can be liberal or conservative. After a brief description of heteroskedasticity and its effects on inference in OLS regression, we discuss a family of heteroskedasticity-consistent standard error estimators for OLS regression and argue investigators should routinely use one of these estimators when conducting hypothesis tests using OLS regression. To facilitate the adoption of this recommendation, we provide easy-to-use SPSS and SAS macros to implement the procedures discussed here.

Addressing Moderated Mediation Hypotheses: Theory, Methods, and Prescriptions.

This article provides researchers with a guide to properly construe and conduct analyses of conditional indirect effects, commonly known as moderated mediation effects. We disentangle conflicting definitions of moderated mediation and describe approaches for estimating and testing a variety of hypotheses involving conditional indirect effects. We introduce standard errors for hypothesis testing and construction of confidence intervals in large samples but advocate that researchers use bootstrapping whenever possible. We also describe methods for probing significant conditional indirect effects by employing direct extensions of the simple slopes method and Johnson-Neyman technique for probing significant interactions. Finally, we provide an SPSS macro to facilitate the implementation of the recommended asymptotic and bootstrapping methods. We illustrate the application of these methods with an example drawn from the Michigan Study of Adolescent Life Transitions, showing that the indirect effect of intrinsic student interest on mathematics performance through teacher perceptions of talent is moderated by student math self-concept.

Longitudinal analysis of the relationship between symptoms and quality of life in veterans treated for posttraumatic stress disorder.

This study examined how change in posttraumatic stress disorder (PTSD) symptoms relates to change in quality of life. The sample consisted of 325 male Vietnam veterans with chronic PTSD who participated in a randomized trial of group psychotherapy. Latent growth modeling was used to test for synchronous effects of PTSD symptom change on psychosocial and physical health-related quality of life within the same time period and lagged effects of initial PTSD symptom change on later change in quality of life. PTSD symptoms were associated with reduced quality of life before treatment. There were synchronous effects of symptom change on change in quality of life but no significant lagged effects. Results indicate the importance of measuring quality of life in future investigations of PTSD treatment.

Believing is seeing? Investigating the perceived accuracy of criminal psychological profiles.

This study investigated whether perceptions of criminal psychological profiles are influenced by the identity of the profile's author. Police officers were given a profile they were told was written by either a professional profiler or by an unspecified author. When judged in relation to the actual perpetrator of the crime, police officers tended to perceive greater accuracy in a profile when it was labeled as authored by a professional profiler independent of the actual content of the profile. But officers' judgments of the usefulness of the profile were not affected by knowledge of who wrote the profile. Explanations for this result focus on the ambiguous nature of criminal profiles and how this ambiguity enhances the likelihood that beliefs about the validity of profiling can color perceptions of the content of the profile.