Testing mediation in nursing research: beyond Baron and Kenny

Area of Science:

• Statistics
• Psychology
• Nursing Research

Background:

• The Baron and Kenny (1986) method for mediation testing is widely used in nursing research.
• Statistical and methodological concerns have been raised regarding the Baron and Kenny approach.
• This paper addresses the need for more robust mediation testing methods in scientific literature.

Purpose of the Study:

• To critically evaluate alternative statistical approaches for testing mediation.
• To compare the performance of various mediation testing methods.
• To identify superior methods for mediation analysis in research.

Main Methods:

• Review of the Baron and Kenny approach and its identified shortcomings.
• Critical analysis of 17 alternative mediation testing methods categorized into causal steps, difference in coefficients, and product of coefficients.
• Evaluation criteria included Type I error rate control, statistical power, computational ease, and versatility.

Main Results:

• Several methods adequately control the Type I error rate.
• The joint significance test of alpha and beta, asymmetric distribution of products test, and percentile bootstrap method demonstrate high statistical power.
• The joint significance test of alpha and beta is identified as superior due to its ease of computation and versatility.

Conclusions:

• Nurse researchers should consider moving beyond the Baron and Kenny approach for mediation testing.
• Adopting more robust statistical tests will enhance knowledge development in nursing.
• Continued research on mediation models is crucial for advancing scientific understanding.