Week 8 Lecture Recording
Introduction to Mediation
Objective: Understand the mediation role in linking a predictor and outcome variable through a mediator.
Differentiate between mediation and moderation analysis.
Learn to assess mediation significance with the Sobel test.
Design studies utilizing mediation for causality assessments.
Definition of Mediation
Mediation: Involvement of three variables, providing a mechanism linking the predictor (X) to the outcome (Y) through a mediator (M).
Also known as the causal hypothesis, it answers: "What are the mechanisms underlying a causal relationship?"
The mediator acts as an explanatory link, helping to understand how and why X affects Y.
Mediation vs. Moderation
Both involve three variables but answer different questions:
Mediation: How and why does X affect Y?
Moderation: Under what conditions is the effect of X on Y evident?
Example of moderation: Gender moderating the relationship between depression and stress.
Key Concepts in Mediation Analysis
Direct, Indirect, and Total Effects
Direct Effect: Relationship between X and Y after controlling for M.
Indirect Effect: Pathway from X to M (pathway a) and M to Y (pathway b); the interesting part of mediation analysis.
Total Effect: Combination of the direct effect and the indirect effect (a + b).
Example Scenarios of Mediation
Medical intervention ➔ Exercise (M) ➔ Reduced Depression (Y).
Anti-tobacco norms (M) lead to reduced tobacco use (Y).
Screening programs (M) identify early-stage cancer ➔ reduced cancer death (Y).
Theoretical Framework
Base concept from Woodworth's stimulus-organism-response theory: The organism stands as the mediator, influencing how a stimulus results in a response.
Importance of identifying mediators causally related to the outcome.
Identifying Mediation Through Statistical Analysis
Baron and Kenny's Approach
Establish mediation using three regression models:
Predict the outcome (Y) from the predictor (X).
Predict the mediator (M) from the predictor (X).
Predict the outcome (Y) from both the mediator (M) and predictor (X).
Conditions for mediation:
Predictor (X) must significantly predict outcome (Y).
Predictor (X) must significantly predict mediator (M).
Mediator (M) must significantly predict outcome (Y).
The relationship between X and Y must diminish when M is included.
Limitations of Baron and Kenny's Approach
Difficulties in defining the extent of relationship reduction necessary for mediation inference.
The Sobel Test
An alternative method to estimate the significance of the indirect effect.
Involves a different equation but remains reliant on p-values.
Addresses sample size influence impacting significant findings.
Practical Example: Pornography Consumption and Infidelity
Mediation model:
Pornography consumption ➔ Relationship commitment (M) ➔ Infidelity (Y).
Mediation analysis involves using SPSS's PROCESS tool to evaluate:
Indirect and direct pathways.
Importance of interpreting p-values and confidence intervals accurately.
Reporting Mediation Analysis
Mediation analysis can be reported in various formats including:
Complete models with total and direct effects.
Diagrams showing beta values and p-values across pathways.
Emphasis on clear reporting for understanding the complete mediation pathways.
Conclusion
This session has covered the foundations of mediation analysis, its statistical assessment, and its application within research. Students will practice these concepts in the upcoming computer labs for real-world application and deeper understanding.