MR

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:

    1. Predict the outcome (Y) from the predictor (X).

    2. Predict the mediator (M) from the predictor (X).

    3. 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.