Mediation Analysis Procedure

Introduction

• Walkthrough for conducting mediation analysis
• Assumes prior exposure to concepts and terms related to mediation
• Steps covered:
  • Estimate total effect (X and Y)
  • Estimate direct effect of X on M
  • Estimate direct effect of X on Y and M on Y
  • Estimate indirect effect
  • Example: Self-oriented perfectionism (X) => Positive effect (Y) via Flourishing (M)

Steps in Mediation Analysis

Step 1: Estimate Total Effect (X and Y)

• Total effect between independent (X) and dependent variable (Y)
• Compare total effect with direct effect in mediation model
• Conduct using bivariate regression (X predicting Y)

Step 2: Estimate Direct Effect of X on M

• Direct effect of independent variable (X) on mediator (M)
• Conduct using bivariate regression (X predicting M)

Step 3A: Estimate Direct Effect of X on Y

• Primary direct effect between independent variable (X) and dependent variable (Y)
• Conduct using multiple regression (X and M predicting Y)
• Obtain unstandardized beta weight

Step 3B: Estimate Direct Effect of M on Y

• Conduct using multiple regression (X and M predicting Y)
• Obtain unstandardized beta weight

Step 4: Test Indirect Effect

• Estimate statistical significance of the indirect effect
• Sobel test or bootstrapping approach

Example: Mediation Analysis (Self-oriented Perfectionism)

Variables

• X: Self-oriented perfectionism
• Y: Positive effect
• M: Flourishing

Conducting the Analysis

• Total Effect (Step 1)
  • Bivariate regression: self-oriented perfectionism => positive effect
  • Coefficients table: Unstandardized beta = 0.102 (significant)
• Direct Effect X on M (Step 2)
  • Bivariate regression: self-oriented perfectionism => flourishing
  • Coefficients table: Unstandardized beta = 0.197 (significant)
  • Values needed for Sobel test: Beta = 0.197, Standard Error = 0.052
• Direct Effects X on Y and M on Y (Steps 3A & 3B)
  • Multiple regression: self-oriented perfectionism + flourishing => positive effect
  • Coefficients: X on Y (Beta = 0.017, SE = 0.030), M on Y (Beta = 0.43, SE = 0.028)
  • X on Y not significant (P = 0.573)
  • Values needed for Sobel test: Betas and SEs
• Statistical Significance of Indirect Effect (Step 4)
  • Sobel test: values for A and B, standard errors
  • Web calculator or SPSS for bootstrapping
  • Example Sobel test result: Z value = 3.68, p < 0.001
  • Sobel test indicates significant mediation (indirect effect = 0.085)

Conclusion from Example

• Self-oriented perfectionism does not directly impact positive effect
• Influence mediated through flourishing
• Indirect effect estimated at 0.085 (significant)

Sobel Test vs. Bootstrapping

• Preference for bootstrapping over Sobel test due to fewer assumptions
• Bootstrapping process using SPSS
  • Requires specific variable names (y_var, x_var, m_var)
  • SPSS syntax to run the test
  • Provides both Sobel test and bootstrapping results

Summary

• Mediation analysis helps to understand the indirect paths through mediators
• Simple in execution: two regressions (bivariate and multiple)
• Importance of testing the significance of indirect effects
• Bootstrapping shown as a robust alternative to Sobel test.