Mediation Analysis Report

Research Question Classification

Question: Does self-efficacy mediate the effect of training on job performance?

Analysis Type: Simple Mediation (PROCESS Model 4)

Rationale: This is a mediation question because it asks whether self-efficacy EXPLAINS or TRANSMITS the effect of training on job performance. The question asks “how” training affects performance (via self-efficacy), not “when” or “for whom.”

Conceptual Path Diagram

           Self-Efficacy (M)
                ↗   ↘
              a       b
             ↗         ↘
    Training (X) ──c'──→ Job Performance (Y)

Path a: Training → Self-Efficacy
Path b: Self-Efficacy → Job Performance (controlling for Training)
Path c': Direct effect of Training on Job Performance (after accounting for M)
Indirect effect: a × b (effect transmitted through Self-Efficacy)
Total effect: c = c' + (a × b)

PROCESS Model Selection

Model 4 - Simple Mediation

This is the appropriate model because:

• One independent variable (Training)
• One mediator (Self-Efficacy)
• One dependent variable (Job Performance)
• No moderators specified
• Testing whether self-efficacy explains the training-performance link

Runnable Code

SPSS PROCESS Macro

PROCESS y=JobPerformance /x=Training /m=SelfEfficacy /model=4
        /boot=5000 /conf=95.

R (mediation package)

library(mediation)

# Fit mediator model (a path)
med_model <- lm(SelfEfficacy ~ Training, data = df)

# Fit outcome model (b and c' paths)
out_model <- lm(JobPerformance ~ Training + SelfEfficacy, data = df)

# Run mediation analysis with bootstrapping
med_result <- mediate(med_model, out_model,
                      treat = "Training", mediator = "SelfEfficacy",
                      boot = TRUE, sims = 5000)

# View results
summary(med_result)

Python (pingouin)

import pingouin as pg
import pandas as pd

result = pg.mediation_analysis(
    data=df,
    x='Training',
    m='SelfEfficacy',
    y='JobPerformance',
    n_boot=5000,
    seed=42
)

print(result)

How to Interpret Results

Critical Effects to Examine

Decision Rules

Mediation is supported if:

• Bootstrap 95% CI for the indirect effect (a×b) does NOT include zero
• Example: CI [0.12, 0.38] → Significant mediation
• Example: CI [-0.05, 0.25] → No significant mediation

Type of mediation:

• Partial mediation: Indirect effect is significant AND direct effect (c’) is significant
• Full mediation: Indirect effect is significant AND direct effect (c’) is non-significant
  • Note: Modern practice avoids “full mediation” terminology; report both effects

APA-Formatted Results Section

To test whether self-efficacy mediated the relationship between training 
and job performance, we conducted a mediation analysis using PROCESS Model 4 
(Hayes, 2022) with 5,000 bootstrap samples.

Results indicated that training significantly predicted self-efficacy 
(a = [INSERT COEF], SE = [INSERT SE], t = [INSERT t], p = [INSERT p]). 
Self-efficacy significantly predicted job performance controlling for 
training (b = [INSERT COEF], SE = [INSERT SE], t = [INSERT t], p = [INSERT p]). 
The direct effect of training on job performance remained significant 
(c' = [INSERT COEF], SE = [INSERT SE], t = [INSERT t], p = [INSERT p]).

Critically, the indirect effect of training on job performance through 
self-efficacy was significant (ab = [INSERT COEF], 95% CI [[INSERT LL], 
[INSERT UL]]), supporting partial mediation. The indirect effect accounted 
for approximately [INSERT %]% of the total effect of training on job 
performance.

Common Pitfalls to Avoid

1. DON’T use p-values for the indirect effect → Use bootstrap CI only
2. DON’T claim causality with cross-sectional data → State limitations clearly
3. DON’T say “fully mediated” if c’ is significant → Partial mediation is valid and common
4. DON’T forget to report all path coefficients → Include non-significant paths

Reference

Hayes, A. F. (2022). Introduction to mediation, moderation, and conditional process analysis: A regression-based approach (3rd ed.). Guilford Press.