Mediation and Moderation

Analysis

https://github.com/davidfoxcrof
t/lsj-data/tree/master/data
• https://lsj.readthedocs.io/en/latest/
• https://youtu.be/dv5CGzD0hXM
• https://library.virginia.edu/data/articles/introduction-to-mediation-
analysis
• https://apastyle.apa.org/style-grammar-guidelines/tables-
figures/sample-tables#correlation
• https://library.virginia.edu/data/articles?page=2
Mediator definitions
• A mediator is a variable in a chain whereby an independent variable causes
the mediator which in turn causes the outcome variable (Sobel, 1990)
• A variable that occurs in a causal pathway from independent to dependent
variables. It causes variation in the dependent variable and itself is caused
to vary by the independent variable (Last, 1988).
• Mediation analysis tests a hypothetical causal chain where one
variable X affects a second variable M and, in turn, that variable
affects a third variable Y. Mediators describe the how or why of a
(typically well-established) relationship between two other variables
and are sometimes called intermediary variables since they often
describe the process through which an effect occurs. This is also
sometimes called an indirect effect.
Other names for Mediators and the Mediated Effect

• Intervening variable is a variable that comes in between two others.

• Process variable because it represents the process by which X affects
Y.
• Indirect Effect for Mediated Effect to indicate that there is a direct
effect of X on Y and there is an indirect effect of X on Y through M.
Mediation Analysis
• A variable M is intermediate in the causal sequence relating X to Y.
Ex:
1. Psychotherapy induces catharsis, insight, and other mediators
which lead to a better outcome (Freedheim & Russ, 1992)
2. Psychotherapy changes attributional style which reduces
depression (Hollon, Evans, & DeRubies, 1990)
3. Parenting programs reduce parents’ negative discipline which
reduces symptoms among children with ADHD (Hinshaw, 2002).
4. Positive Parenting (X) of an infant predicts selfesteem (M) which
predicts positive parenting as an adult (Y).
Mediation Analysis importance
• Central questions in many fields are about the mediating process.
• Important for basic research on mechanisms of effect.
• Critical for applied research, especially prevention and treatment.
Mediation Assumptions
1.The dependent, independent, and mediator variables (the
variables of interest) need to be using a continuous scale.
2.The variables of interest (the dependent variable and the
independent and mediator variables) should have a linear
relationship, which you can check with a scatterplot
3.The data must not show multicollinearity (see Multiple
Regression).
4.There should be no spurious outliers, and the distribution of
the variables should be approximately normal.
Higher grades predict higher happiness: X (grades) → Y (happiness)
• If good grades boost one’s self-esteem and then high self-esteem
boosts one’s happiness:
• X (grades) → M (self-esteem) → Y (happiness).
How to analyze mediation effects
• As any other regression analysis, mediation analysis does not imply
causal relationships unless based on experimental design.
• To analyze mediation:
1.Follow Baron & Kenny’s steps
2.Use either the Sobel test or bootstrapping for significance testing.
To analyze mediation: Follow Baron & Kenny’s steps

• Use either the Sobel test or bootstrapping for significance testing.

• A mediation analysis is comprised of three sets of regression: X → Y, X

→ M, and X + M → Y.

The regression coefficient for the indirect effect represents the

change in Y for every unit change in X that is mediated by M.
There are two ways to estimate the indirect coefficient. Judd and
Kenny (1981) suggested computing the difference between two
regression coefficients. To do this, two regressions are required.
Step 1.
Is 𝑏1 significant? We want X to affect Y. If there is no
relationship between X and Y, there is nothing to mediate.

Although this is what Baron and Kenny originally

suggested, this step is controversial.
Even if we don’t find a significant association between X
and Y, we could move forward to the next step if we have
a good theoretical background about their relationship.
Step 2.
Step 3.
• If the effect of X on Y completely disappears, M fully mediates between X and Y (full
mediation).

• If the effect of X on Y still exists, but in a smaller magnitude, M partially mediates

between X and Y (partial mediation).

• The example shows a full mediation, yet a full mediation rarely happens in practice.

• once we find these relationships, we want to see if this mediation effect is statistically
significant (different from zero or not).

• To do so, there are two main approaches:

• The Sobel test (Sobel, 1982) and bootstrapping (Preacher & Hayes, 2004).
Mediate () takes two model objects as input (X → M and X + M → Y) and we need
to specify which variable is an IV (treatment) and a mediator (mediator).
For bootstrapping, set boot =to at least 500.
Mediation Estimates
95% Confidence
Interval
%
Effect Label Estimate SE Lower Upper Z p
Mediation
Indire 0.356 0.081
a×b 0.197 0.516 4.382 < .001 90.00
ct 5 4
0.039 0.108 -
Direct c 0.251 0.367 0.714 10.00
6 0 0.172
c+a 0.396 0.110 100.0
Total 0.180 0.612 3.600 < .001
×b 1 0 0
By the way, we don’t have to follow all three steps as Baron and Kenny suggested. We could run two
regressions (X → M and X + M → Y) and test its significance using the two models. However, the suggested
steps help you understand how it works!

Path Estimates
95% Confidence
Interval
Label Estimate SE Lower Upper Z p
0.56 0.09 0.37 0.74 5.99 < .00
X → M a
10 35 8 4 8 1
0.63 0.09 0.44 0.83 6.41 < .00
M → Y b
55 90 1 0 8 1
-
0.03 0.10 0.25 0.36 0.71
X → Y c 0.17
96 80 1 7 4
2
• Mediation analysis is not limited to linear regression; we can use
logistic regression or polynomial regression and more.
• Also, we can add more variables and relationships, for example,
moderated mediation or mediated moderation.
• However, if your model is very complex and cannot be expressed as a
small set of regressions, you might want to consider structural
equation modeling instead.
To sum up, here’s a flowchart for mediation analysis
Bootstrapping
• Bootstrapping is a statistical method that utilizes random resampling
with replacement to estimate a population parameter.
• This technique samples from a given dataset to estimate a parameter
when it would otherwise be impossible or impractical to do so.
• Those samples are used to calculate standard errors, confidence
intervals and for hypothesis testing.
Direct Effect, Indirect Effect, And Total Effects
Forms of Mediation
Forms of Mediation
Forms of Mediation
• Complementary mediation is where the direct and indirect effects have a similar
influence regarding directionality. For instance, the direct effect may a have
positive influence, and the indirect effect has a positive influence as well.

• Competitive mediation is where you have different directionality between direct

and indirect effects. The direct effect might have a negative influence, but the
indirect effect might have a positive influence. With this type of mediation, the
mediator's presence can change the influence's directionality.
Mediator versus Moderator

• Moderator is a variable that affects the strength of the relation

between two variables. The variable is not intermediate in the causal
sequence, so it is not a mediator, but it could be in a causal sequence.

• Moderator is usually an interaction, the relation between X and Y

depends on a third variable.
Moderation analysis
• Moderation describes a situation in which the relationship between two constructs is not
constant but depends on the values of a third variable, referred to as a moderator variable.
• The moderator variable (or construct) changes the strength, or even the direction of a
relationship between two constructs in a model.
• For example, prior research has shown that the relationship between customer satisfaction and
customer loyalty differs as a function of the customers’ income or age (e.g., Homburg & Giering,
2001).
• More precisely, income has a pronounced negative effect on the satisfaction-to-loyalty
relationship – the higher the income, the weaker the relationship between satisfaction and
loyalty.
• In short, income serves as a moderator variable that accounts for heterogeneity in the data.
• This means the satisfaction-to-loyalty relationship is not the same for all customers but differs
depending on the income level.
• In this respect, moderation can (and should) be seen as a means to account for heterogeneity in
the data
Moderation Model
Example

• whether the relationship between the number of hours of sleep (X) a

graduate student receives and the attention that they pay to this tutorial

(Y) is influenced by their consumption of coffee (Z).

• Here we create the moderation effect by making our DV (Y) the

product of levels of the IV (X) and our moderator (Z).

Moderator analysis
• A moderator analysis determines whether the relationship between two
variables depends on (is moderated by) the value of a third variable.

• This relationship is commonly between (a) a continuous dependent variable

and a continuous independent variable, which is modified by a dichotomous
moderator variable.

• a moderator analysis is really just a multiple regression equation with an

interaction term.
Variation of moderation

(a) A continuous dependent variable and continuous independent

variable, which is modified by a dichotomous moderator variable

(b) A continuous dependent variable and continuous independent

variable, which is modified by a polytomous moderator variable; or

(c) A continuous dependent variable and continuous independent

variable, which is modified by a continuous moderator variable.
Assumptions of a moderator analysis with a
dichotomous moderator.
1: DV should be measured on a continuous scale (i.e., an interval or ratio variable).
2: One IV, which is continuous (i.e., an interval or ratio variable) and one moderator
variable that is dichotomous (i.e., a nominal variable with two groups).
3: You should have independence of observations (i.e., independence of
residuals), which you can check using the Durbin-Watson statistic.
4: There needs to be a linear relationship between the dependent variable and the
independent variable for each group of the dichotomous moderator variable. This
can be checked by creating a scatterplot
5: Data needs to show homoscedasticity, which is when the error variances are the
same for all combinations of independent and moderator variables.
Assumptions
• 6: Data must not show multicollinearity, which occurs when you have two or more independent
variables that are highly correlated with each other. This leads to problems with understanding
which independent variable contributes to the variance explained in the dependent variable, as well
as technical issues in calculating a multiple regression model (to detect multicollinearity through an
inspection of correlation coefficients and Tolerance/VIF values)

• 7: There should be no significant outliers, high leverage or highly influential points. Outliers,
leverage , and influential points are different terms used to represent observations in data sets that
are in some way unusual when you wish to perform a moderator analysis.
Assumptions

• 8: Check that the residuals (errors) are approximately normally distributed.

• Methods to do this can be based either on graphical or numerical methods (The

Shapiro-Wilk test for normality).

Moderation data set
• https://www.tandfonline.com/doi/full/10.1080/00224545.2018.1523
783
• http://hdl.handle.net/2374.MIA/6254
• Mindful multitasking: Disentangling the effect of
polychronicity on work–home conflict and life satisfaction

Polychronicity refers to the preference for dealing with multiple tasks or activities
simultaneously, as opposed to sequentially.

IV: Polychronicity
Moderator: mindful multitasking
DV: life satisfaction
outputs

Moderation Estimates
95% Confidence Interval
Estimate SE Lower Upper Z p
Poly_Ct -0.1286 0.0848 -0.2948 0.0375 -1.518 0.129
MAAS_C
0.0635 0.0784 -0.0902 0.2171 0.809 0.418
t
Poly_Ct
✻
0.2016 0.0730 0.0586 0.3446 2.763 0.006
MAAS_C
t
interpretation
1.Main Effect of Polychronicity (Poly_Ct):
1. The main effect of Polychronicity (Poly_Ct) is not statistically significant (p = 0.129). This
suggests that, on its own, Polychronicity does not significantly predict the outcome variable.
2.Main Effect of Mindfulness (MAAS_Ct):
1. The main effect of Mindfulness (MAAS_Ct) is also not statistically significant (p = 0.418). This
indicates that, on its own, Mindfulness does not significantly predict the outcome variable.
3.Interaction Effect (Poly_Ct ✻ MAAS_Ct):
1. The interaction effect between Polychronicity and Mindfulness (Poly_Ct ✻ MAAS_Ct) is
statistically significant (p = 0.006).
2. A positive coefficient (0.2016) indicates that the relationship between Polychronicity and the
outcome variable strengthens as levels of Mindfulness increase.
3. This suggests that individuals with higher levels of Mindfulness demonstrate a stronger
association between their preference for dealing with multiple tasks simultaneously
(Polychronicity) and the outcome variable compared to individuals with lower levels of
Mindfulness.
interpretation
• The moderation analysis results indicate that the interaction effect
between Polychronicity (Poly_Ct) and Mindful Attention and
Awareness Scale (MAAS_Ct) on the outcome variable is statistically
significant, B = 0.20, SE = 0.07, 95% CI [0.06, 0.34], z = 2.76, p = 0.006.
• This suggests that the relationship between Polychronicity and the
outcome variable is moderated by levels of mindfulness. Specifically,
as levels of mindfulness (MAAS_Ct) increase, the association between
Polychronicity and the outcome becomes stronger.
The simple slope estimates provided indicate how the predictor variable (Poly_Ct: Polychronicity)
predicts the dependent variable (LifeSat: Life Satisfaction) at different levels of the moderator
variable (MAAS_Ct: Mindful Attention and Awareness Scale).

Simple Slope Estimates

95% Confidence Interval
Estimate SE Lower Upper Z p
Average -0.129 0.0873 -0.2998 0.0425 -1.47 0.141
Low (-
-0.376 0.1491 -0.6683 -0.0837 -2.52 0.012
1SD)
High
0.119 0.0974 -0.0722 0.3097 1.22 0.223
(+1SD)
Note. shows the effect of the predictor (Poly_Ct) on the dependent variable (LifeSat) at different levels of the
moderator (MAAS_Ct)
slope
• At the average level of mindfulness, the effect of Polychronicity on Life Satisfaction is not
statistically significant (p = 0.141). This suggests that, on average, there is no significant
relationship between Polychronicity and Life Satisfaction when considering individuals with
average levels of mindfulness.
• At low levels of mindfulness (1 standard deviation below the mean), there is a statistically
significant negative relationship between Polychronicity and Life Satisfaction (p = 0.012). This
indicates that individuals with lower levels of mindfulness tend to experience lower life satisfaction
as their preference for dealing with multiple tasks simultaneously (Polychronicity) increases.
• At high levels of mindfulness (1 standard deviation above the mean), the effect of Polychronicity
on Life Satisfaction is not statistically significant (p = 0.223). This suggests that individuals with
higher levels of mindfulness do not show a significant relationship between Polychronicity and
Life Satisfaction.
• Specifically, the relationship is significant at low levels of mindfulness,
indicating that individuals with lower mindfulness experience a
decrease in life satisfaction as their preference for multitasking
increases. However, at average and high levels of mindfulness, the
relationship between Polychronicity and Life Satisfaction is not
significant, suggesting that mindfulness moderates this relationship.
Abstract of the study
• This study seeks to disentangle the effect of polychronicity on
work–home conflict, home–work conflict, and life satisfaction, by
evaluating mindfulness as a moderator. We propose that
mindfulness moderates the relationship between polychronicity
and work–home and home–work conflict such that the
relationship will be negative when mindfulness is high and
positive when mindfulness is low. Additionally, we propose that
mindfulness moderates the relationship between polychronicity
and life satisfaction such that the relationship will be positive
when mindfulness is high and negative when mindfulness is low.
A total of 138 academics throughout India completed measures
of polychronicity, mindfulness, life satisfaction, and work-to-home
and home-to-work conflict scales. The moderation findings
illustrate that higher levels of mindfulness enhance the effects of
polychronicity.