SPSS Statistical Computing Workshop

Moderation and Mediation using the Process Macro

This workshop is intended to introduce the use of the Process macro for
SPSS. The Process macro was written by Andrew F. Hayes.

Some Terms
Moderation: Moderation refers to an interaction. A moderator variable is a
variable that interacts with another variable, often with an independent or
predictor variable. If two variables A and B interact, then the effect of A
depends on the value of B.
Mediation: In mediation the effect of one variable is transmitted to another
variable through the mediator variable. For example the effect of X is
transmitted to Y through variable M.
This can be seen in the diagram below of a classical mediation model
(adapted from Baron & Kenny, 1986). In the diagram "a" is the regression
coefficient predicting M by X. "b" is the coefficient predicting Y by M, And,
"c'" is the coefficient predicting Y from X.

Partitioning the Effect of X on Y: In a mediation model the effect of

variable X on Y can be partitioned into two parts: 1) the direct effect of X on
Y, and 2) the indirect effect of X on Y via the mediator M. Combined, the
direct and indirect effect of X on Y is known as the total effect.
Direct Effect: The direct effect is the effect of X on Y when the mediator is
included in the model. In the diagram above the direct effect is shown as
"c'."
Indirect Effect: The indirect effect is a measure of how much of the effect
of X on Y that is being mediated. Another term for the indirect effect is the
mediation effect.
In a classical mediation model the indirect effect is obtained by multiplying
the "a" coefficient times the "b" coefficient in the diagram above.
Conditional Indirect Effects: When the size of an indirect effect depends
on the values of another variable, we call this a conditional indirect effect.
Basically, there is an interaction in the model that affects the indirect effect.
Collectively, we refer to these types of models as moderated mediation.

About the Process Macro

The Process macro will compute models with continuous or binary outcome
variables. The primary predictor variable (IV) can be continuous or binary.
However, the mediator variable must be continuous. The macro will not
work correctly if the primary predictor is categorical and will not run if the
mediation variable is categorical.

How to get and install the Process macro

1) Go to Andrew Hayes' webpage, http://afhayes.com/spss-sas-and-mplusmacros-and-code.html
2) Download the process macro.
3) Paste the macro into a syntax window.
4) Run the macro from the syntax window.

Using the Process macro

There are two ways to use the Process macro. Process can be used with
point-and-click module that needs to be downloaded or it can be used
directly from the SPSS syntax window. We will demonstrate the Process
macro using the syntax method.
You begin by entering the command "process" followed by the keyword
"vars =". This is where you list all of the variables to be used in the
analysis. The keywords "y", "x", and "m" are used for the response variable,
the predictor variable and the mediator variable, respectively. The
moderator variable for moderated mediation will be "w" or "v" depending on
the model. Finally, it is very important to include the model number after the
"model = ".
The model number can be found in an appendix of Hayes' book or in a pdf
file of model templates found on the author's website.
Process does not output a t-statistic or p-value for the indirect effect.
Rather, Process bootstraps (resamples) the indirect effect and outputs a
95% confidence interval. A hypothesis can be made by determining
whether zero falls inside the confidence interval. If the interval includes
zero then the indirect effect is not significant at the 0.05 level. If zero is not
in the interval then the indirect effect is statistically significant at the 0.05
level.

The data
We will demonstrate the Process macro using
the mediation_data.sav dataset. The file has five continuous standardized
test scores; read, write, math, science and socst. It also has a binary
variable, hisci, which is an indicator of a high science score. With the
exception of the first example, the response variable with be

either science or hisci. Again, except for the first example, we will
use math as the predictor variable and read as the mediator.

Example 1: Moderation
In this example, there is an interaction between the predictor, X and the
moderator, M. For this example, the dependent or response variable
is read; the independent or predictor variable is socst; and the moderator
is math. Here is a block diagram of a model with a moderator variable.

* Example 1 (Process model 1).

process vars = socst read math
/ y = read
/ x = socst
/ m = math
/ model = 1.
Example 2: Classical Mediation
In this example, the DV is science and the IV is math; read is the
mediator. Here is a block diagram of a simple classical mediation.

* Example 2 (Process model 4).

process vars = science read math
/ y = science
/ x = math
/ m = read
/ model = 4.
Example 3: Moderated Mediation
This example has the same variables as the previous model with the
addition of an interaction between the IV and the mediator. Here is a block
diagram of the model.

* Example 3 (Process model 74).

process vars = science read math
/ y = science
/ x = math
/ m = read
/ model = 74.
Example 4: Moderated Mediation
This example has the same variables as the previous model with the
addition of an interaction between the IV and the mediator. Here is a block
diagram of the model.

* Example 4 (Process model 8).

process vars = science read math write
/ y = science
/ x = math
/ m = read
/ w = write
/ model = 8.
Example 5: Moderated Mediation
This time there is an interaction between the MV and the mediator. Here is
a block diagram of the model.

* Example 5 (Process model 14).

process vars = science read math write
/ y = science
/ x = math
/ m = read
/ v = write
/ model = 14.

Example 6: Simple Mediation with Binary DV

In this example, the DV and IV are the same; and read is the mediator. The
block diagram is the same as example 2.

* Example 6 (Process model 4).

process vars = hisci read math
/ y = hisci
/ x = math
/ m = read
/ model = 4.
Example 7: Moderated Mediation with Binary DV
This example has the same variables as the previous model with the
addition of an interaction between the IV and the mediator. The block
diagram is the same as for Example 3.

* Example 7 (Process model 74).

process vars = hisci read math write
/ y = hisci
/ x = math
/ m = read
/ model = 74.
Example 8: Moderated Mediation with Binary DV
This example has the same variables as the previous model with the
addition of an interaction between the IV and the mediator. The block
diagram is the same as for Example 4.

* Example 8 (Process model 8).

process vars = hisci read math write
/ y = hisci
/ x = math

/ m = read
/ w = write
/ model = 8.
Example 9: Moderated Mediation with Binary DV
This example has the same variables as the previous model with the
addition of an interaction between the IV and the mediator. The block
diagram is the same as for Example 5.

* Example 9 (Process model 14).

process vars = hisci read math write
/ y = hisci
/ x = math
/ m = read
/ v = write
/ model = 14.
Beyond Process
The Process macro computes the indirect effect by calculating the product
of coefficients. This method works well when the mediator and response
variables are continuous. However, when the mediator is binary or the
response variable is a count variable (Poisson or negative binomial) the
product of coefficients approach does not work. In general, the product of
coefficients method does not work with nonlinear models.
An alternative method is to use a causal mediation approach which
involves the use of counterfactuals. Programs that do causal mediation can
be found in R, SAS, Stata and Mplus. At the current time there are no
macros for causal mediation in SPSS.
Additionally, the Process macro cannot be used with multilevel models.
Multilevel mediation is a much more complex analysis which goes beyond
the scope of this workshop.