R squared effect-size measures and overlap between direct and indirect effect in mediation analysis.

Area of Science:

• Psychological research methods
• Statistical modeling

Background:

• Mediation analysis is crucial for understanding indirect effects in psychological research.
• A previously published method (Fairchild et al., 2009) aimed to partition variance into direct and indirect effects.

Purpose of the Study:

• To critically evaluate the Fairchild et al. (2009) method for computing variance in mediation analysis.
• To explain the counterintuitive results observed when applying this method.

Main Methods:

• Analysis of the Fairchild et al. (2009) method's handling of the interdependence between direct and indirect effects.
• Discussion of alternative approaches for managing overlap variance in mediation analysis.

Main Results:

• The Fairchild et al. method assigns all overlap variance to the indirect effect, leading to potentially misleading interpretations.
• Situations exist where a strong direct effect is overshadowed by a seemingly larger indirect effect due to this variance assignment.

Conclusions:

• The Fairchild et al. method's approach to overlap variance is problematic and can produce counterintuitive results in mediation analysis.
• Current methods for handling overlap variance in mediation analysis have inherent disadvantages and require careful consideration.