Mediation

Analysis
Psych 203
Quantitative Research in Psychology
Topic #6

Dr. Marshall N. Valencia

[email protected]
 At the end of the session, students should be able to:
Learning • Define and differentiate mediation from moderation
analysis
Objectives • Run mediation analysis using Jamovi
• Interpret outputs and write results of the analysis.

Psych 203 (Quantitative Research in Psychology)

Marshall N. Valencia ([email protected])
 • In predicting the link between an IV and DV, we are expected to generate
Theoretical Explanations one or more theories that could account for the hypothesis
in Psychology • Most psychological “theories” are in the form of either: MEDIATION or
MODERATION
(McGuire, 2004)

Theory of Planned Behavior

Psych 203 (Quantitative Research in Psychology)

Marshall N. Valencia ([email protected])
 Common Types of Typologies
01
Theoretical Frameworks
02 Simple Chains

03 Inventory of Causes

04 Inventory of Effects

Combination of
05
Causes and Effects

Cause and Effect Models

06
with Intervening Variables

Psych 203 (Quantitative Research in Psychology)

Marshall N. Valencia ([email protected])
01 Typologies

02 Simple Chains
Org’l
Inventory of
Commitment
03 Causes

Inventory of Effects
Three-component model of organizational
04 commitment (Meyer and Allen's , 1991)

Combination of
05 Causes and Effects
Hofstede’s Dimensions of Culture
Cause and Effect Models (Hofstede, 1980)
06 with Intervening Variables

Psych 203 (Quantitative Research in Psychology)

Marshall N. Valencia ([email protected])
 Typologies
W X Y Z
01

02 Simple Chains

03 Inventory of Causes Theory of Cultural Adjustment

(Oberg, 1960)
04 Inventory of Effects

Combination of
05 Causes and Effects

Cause and Effect Models

06 with Intervening Variables

Value–belief–norm (VBN) Theory

(Stern et al, 1999)
Psych 203 (Quantitative Research in Psychology)
Marshall N. Valencia ([email protected])
 X1
01 Typologies

02 Simple Chains
X2 Y
03 Inventory of Causes
X3 Y1
04 Inventory of Effects

05 Combination of X Y2
Causes and Effects

Cause and Effect Models

06 with Intervening Variables Y3

Psych 203 (Quantitative Research in Psychology)

Marshall N. Valencia ([email protected])
 X1 Z1

X2 Y Z2

01 Typologies X3 Z3

02 Simple Chains

03 Inventory of Causes

04 Inventory of Effects

Combination of
05 Causes and Effects

Cause and Effect Models

06 with Intervening Variables

Job-Characteristics Model (JCM) by Hackman &

Oldham (1976)

Psych 203 (Quantitative Research in Psychology)

Marshall N. Valencia ([email protected])
 Job demands-resources model of burnout (Demerouti
et al, 2001)

X1

01 Typologies
X2 I1

X3
02 Simple Chains
Y
03 Inventory of Causes
X1

04 Inventory of Effects
X2 I2

Combination of X3
05 Causes and Effects (Ajzen, 1991)

06 Cause and Effect Models with

Intervening Variables
Models for Prediction

r
X Y

Simple regression MODEL

• T-test (when X is dichotomous)
• ANOVA (when X has more than 2
categories)
• Linear Regression (when X is continuous)

Psych 203 (Quantitative Research in Psychology)

Marshall N. Valencia ([email protected])
Models for Prediction β
X1
r Y
Multiple regression MODEL
X2 β

• Factorial ANOVA
(when X1 and X2 are categorical)
• Multiple linear Regression
(when X1 and X2 are continuous, or dichotomous)

Psych 203 (Quantitative Research in Psychology)

Marshall N. Valencia ([email protected])
Models for Prediction
X1 β

r
β
r
X2 Y
r β

X3

Multiple regression MODEL

Psych 203 (Quantitative Research in Psychology)
Marshall N. Valencia ([email protected])
Models for Prediction

M Z

X Y X Y
Simple Mediation Model Simple Moderation Model

• Series of Multiple Regression Analysis

Psych 203 (Quantitative Research in Psychology)

Marshall N. Valencia ([email protected])
Models for Prediction
M1 Z

X Y X Y

M2 w

• Complex versions of mediation and moderation models

Psych 203 (Quantitative Research in Psychology)

Marshall N. Valencia ([email protected])
Models for Prediction

Structural Equations Modeling

Psych 203 (Quantitative Research in Psychology)

Marshall N. Valencia ([email protected])
 Mediation Moderation
Differentiating Mediation
from Moderation
• A variable is a mediator (M) • A third variable (Z) is said to
between a predictor (X) and an moderate the relationship
outcome (Y) if the predictor between two other variables (X
variable first has an effect on and Y) if the degree of
the mediator variable, and this relationship between X and Y is
in turn influences the outcome affected by the level of Z
variable
• The mediator is called an
intervening or process variable

X M Y X Y

Psych 203 (Quantitative Research in Psychology)

Marshall N. Valencia ([email protected])
 Mediation Moderation
Differentiating Mediation
Z
from Moderation
X M Y
X Y
Workplace Stress
Conflict Job Performance Sensitivity

Workplace Stress
Conflict

“because of….”
• Workplace conflict will affect job
“especially if”
performance “because of” stress. • Workplace conflict leads to stress,
• Workplace conflict leads to stress which in “especially if” employees are high in
turn affects job performance sensitivity.
Formal hypothesis: Formal hypothesis:
• Stress mediates the relationship between • Sensitivity moderates the relationship
workplace conflict and job performance. between workplace conflict and
• Workplace conflict is positively correlated stress.
with stress; stress is negatively correlated • Sensitivity interacts with workplace
with job performance conflict leading to stress
 Mediation Moderation
Differentiating Mediation
from Moderation
• Tend to change in relation to • Relatively immune to change
other variables / process over time (gender, individual
variables (e.g. stress, differences, etc.)
commitment, self-efficacy) • Answers “when” question
• Answers “why” question • Similar to “interaction” effects
in ANOVA

X M Y X Y

Psych 203 (Quantitative Research in Psychology)

Marshall N. Valencia ([email protected])
Mediation Analysis
M
Well
Being

X Y
Career Job Satisfaction
Growth

• Answers HOW questions; mechanisms which explains the

relationship between X and Y
• 21st century practice of mediation analysis no longer imposes
evidence of simple association between X and Y as a precondition

Psych 203 (Quantitative Research in Psychology)

Marshall N. Valencia ([email protected])
Mediation Analysis Testing A Simple Mediation Model

Steps:
1. Estimate the direct and indirect effects (thru a series
of regression analysis)
2. Statistical Inference (test the significance of the
indirect effect)

Psych 203 (Quantitative Research in Psychology)

Marshall N. Valencia ([email protected])
 M
Step 1: a Well
b
Estimate the direct and indirect
Being
effects (thru a series of
regression analysis) X Y
Career
c
Job Satisfaction
Growth

• Direct effect = path c

• Indirect effect (or mediation effect) = (path a) x (path b)

Effects can be estimated using two regression equations:

Y=constant + bM+ cX
M=constant + aX

Psych 203 (Quantitative Research in Psychology)

Marshall N. Valencia ([email protected])
 M
a Well b
Being
X Y
Career
c
Job Satisfaction Y=constant + bM+ cX
Growth

M=constant + aX

Indirect effect (or mediation effect) = (path a) x (path b)

Psych 203 (Quantitative Research in Psychology)
Marshall N. Valencia ([email protected])
Step 2: Statistical Inference • Inference test for the direct effect
(test the significance of the 1. Null hypothesis Test
indirect effect) 2. Interval Estimation
• Inference test for the indirect effect
1. Normal Theory approach (also known as Sobel Test)
2. Bootstrap Confidence Interval
3. Bias-corrected bootstrap
4. Monte Carlo
5. Distribution of the product

Psych 203 (Quantitative Research in Psychology)

Marshall N. Valencia ([email protected])
Bootstrap Confidence Interval
• Interpretation: if the interval from the lower limit confidence interval (LLCI) to
the upper limit does not contain the value “0,” then the effect is statistically
significant.

Psych 203 (Quantitative Research in Psychology)

Marshall N. Valencia ([email protected])
 M
.278*** .316***
Sample Write-up Well
Being

X Y
.489***
Career Job Satisfaction
Growth

Indirect effect = .088, bootstrap S.E.=.0251, BCI: .0436 to .138

A simple mediation analysis was conducted using ordinary least squares path
analysis to test if well-being mediates the relationship between career growth and
job satisfaction. As can be seen in Figure 1, career growth perceptions is positively
related with well being (β=.278, p<.001) and well-being is in turn positively related
with job satisfaction (β=.316, p<.001). Career growth perceptions affect job
satisfaction directly (β=.489, p<.001) and indirectly through well-being. The bias-
corrected bootstrap confidence interval for the indirect effect (β=.088) based on
5,000 bootstrap samples was entirely above zero (0439 to .1421).

Psych 203 (Quantitative Research in Psychology)

Marshall N. Valencia ([email protected])
Jamovi procedures for
simple mediation
• Make sure Advanced Mediation
Models is installed

• Move the corresponding dependent, mediator, factor, and covariate variables

• In “Mediation options” → click Bootstrap (Percent)

Psych 203 (Quantitative Research in Psychology)

Marshall N. Valencia ([email protected])
 Jamovi procedures for Multiple Mediators (parallel)
parallel mediation

M1
X Commitment Y
(Affective)
Career Job
Growth Satisfaction

M2
Well-being

Psych 203 (Quantitative Research in Psychology)

Marshall N. Valencia ([email protected])
 Jamovi procedures for Multiple Mediators (serial)
serial mediation

X M1 M2 Y

Career Commitment Job

Move the corresponding (Affective)
Well-being
Growth Satisfaction
dependent, mediator, factor,
and covariate variables
In “Mediation options” → click
Bootstrap (Percent)
In “Mediator Models” ….

Psych 203 (Quantitative Research in Psychology)

Marshall N. Valencia ([email protected])
 Exercise
Come up with an intuitively logical hypothesis (based on the
available variables in the dataset: JDRmodel.sav) that reflects:
1. A simple mediation model
2. A parallel mediation model
3. A serial mediation model
For every hypothesis:
1. State the research problem (s)
2. Test the model
3. Summarize the results (use the worksheet)
4. Prepare the write-up.
 Exercise (use JDRmodel.sav)
Variables Variables
Performance Career Growth
Burnout Supervisor Support
Job Satisfaction Social Support
Well-being Teamwork
Engagement Trust in Supervisor
Turnover Intention Trust in Management
Trust in Co-worker
Role Clarity
Voice
Recognition
Optimism Work load
Resiliency Work-home conflict
Hope Work-life conflict
Core Self-Evaluation
Sex, Age, Position, Tenure, Civil status
No. of Children, Education
 • Field, A. (2013). Discovering statistics using SPSS, 4th
Ed. London: Sage Publications.
References • Hayes, A.F. (2018). Introduction to Mediation,
Moderation, and Conditional Process Analysis: A
Regression-Based Approach. USA: Guilford Press

Psych 203 (Quantitative Research in Psychology)

Marshall N. Valencia ([email protected])