Confidence Limits for the Indirect Effect: Distribution of the Product and Resampling Methods

Abstract

The most commonly used method to test an indirect effect is to divide the estimate of the indirect effect by its standard error and compare the resulting z statistic with a critical value from the standard normal distribution. Confidence limits for the indirect effect are also typically based on critical values from the standard normal distribution. This article uses a simulation study to demonstrate that confidence limits are imbalanced because the distribution of the indirect effect is normal only in special cases. Two alternatives for improving the performance of confidence limits for the indirect effect are evaluated: (a) a method based on the distribution of the product of two normal random variables, and (b) resampling methods. In Study 1, confidence limits based on the distribution of the product are more accurate than methods based on an assumed normal distribution but confidence limits are still imbalanced. Study 2 demonstrates that more accurate confidence limits are obtained using resampling methods, with the bias-corrected bootstrap the best method overall.

Peritraumatic dissociation predicts posttraumatic stress disorder symptoms via dysfunctional trauma-related memory among war-affected children

“…For structural equation modelling, full information maximum likelihood estimation with bootstrapped standard errors from 5000 resamples was used. Asymmetric confidence intervals based on bias-corrected bootstrap estimates were constructed for total, indirect, and direct effects to ensure accurate assessment of their significance and take into account the non-normal distribution of the indirect effect (MacKinnon, Lockwood, & Williams, 2004). …”

Internalization as a mediator of the relationship between conformity to masculine norms and body image attitudes and behaviors among young men in Sweden, US, UK, and Australia

“…Mediation effects were tested for significance using Sobel's z test, which relies on the a (IV  MV) and b (MVDV, controlling for IV) paths in the model as well as their standard errors. The biascorrected bootstrap resampling method (Shrout & Bolger, 2002) was implemented with 2000 resamples to more accurately estimate these parameters (MacKinnon, Lockwood, & Williams, 2004). All mediation effects were tested in the same model (Figure 1) rather than running separate mediation models for the two mediators (internalization general and athletic subscales) and the four DVs (desire for leanness, drive for thinness, and attitudinal and behavioral aspects of drive for muscularity), as the latter strategy is likely to over-estimate effect sizes due to model misspecification (Kline, 2005).…”

Effects of a multi-component camp-based intervention on inflammatory markers and adipokines in children: A randomized controlled trial

“…The statistical significance of mediation effects was assessed by Sobel approach (Sobel, 1982). Furthermore, a bootstrapping approach (Mackinnon et al, 2004) (5000 bootstrap resamples) was used to obtain bias-corrected 95% confidence interval (CI) of the mediation effect (a×b) and direct effect (path c'). The proportion of mediation was calculated by dividing the mediation effect (a×b) by the total effect (c).…”