Description

.2 Panduan Model PROCESS

All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.

Related Documents

Share

Transcript

SPSS PROCESS documentation, 2 January 2013 Page 1
PROCESS
PROCESS VARS =
varlist
/Y =
yvar
/X =
xvar
/M =
mvlist
/MODEL =
num
[/W =
wvar
] [/Z =
zvar
] [/V =
vvar
] [/Q =
qvar
] [/WMODVAL =
wval
] [/ZMODVAL =
zval
] [/VMODVAL =
vval
] [/QMODVAL =
qval
] [/MMODVAL =
mval
] [/XMODVAL = xval] [/CLUSTER =
clvar
] [/CONTRAST =
cst
] [/BOOT =
z
{1000**}] [/CONF =
ci
{95**}] [/EFFSIZE =
size
{0**}] [/NORMAL =
t
{0**}] [/JN =
j
{0}] [/COEFFCI = cci{1***}] [/VARORDER =
vord
(2**}] [/HC3 =
h
{0**}] [/COVMY =
cov
{0**}] [/TOTAL =
tot
{0**}] [/CENTER =
cntr
{0**}] [/QUANTILE =
qnt
{0**}] [/DETAIL =
dt
{1**}] [/PLOT = pt{0**}] [/SEED = sd{random**} [/PERCENT = pb{0**}] [/ITERATE = it{10000*}] [/CONVERGE = cvg{.00000001}].
Subcommands in brackets are optional. User input in italics. Brackets should not be included in PROCESS command. ** Default if subcommand is omitted
Overview
PROCESS
is a computational tool for path analysis-based moderation and mediation analysis as well as their combination as a
“
conditional process model
”
(Hayes and Preacher, in press; Hayes, in press). In a single command, it provides many of the capabilities of SOBEL (Preacher and Hayes, 2004), INDIRECT (Preacher & Hayes, 2008), MODPROBE (Hayes & Matthes, 2009), MODMED (Preacher, Rucker, & Hayes, 2007), and MED3/C (Hayes, Preacher, & Myers, 2011) while greatly expanding the
number and complexity of models that combine moderation and mediation (“mediated moderation” and “
moderated mediation
”). In addition to estimating the coefficients of the model
using OLS regression (for continuous outcomes) or maximum likelihood logistic regression (for dichotomous
outcomes
), PROCESS generates direct and indirect effects in mediation and mediated moderation models, conditional effects in moderation models, and conditional indirect effects in moderated mediation models with a single or multiple mediators. PROCESS offers various tools for probing 2 and 3 way interactions and can construct percentile based bootstrap confidence intervals for conditional and unconditional indirect effects. In mediation models, multiple mediator variables can be specified to operate in parallel or in sequence. Heteroscedasticity-consistent standard errors are available for
PROCESS is written by Andrew F. Hayes, http://www.afhayes.com Page 2
inference about paths coefficients, in the Sobel test for indirect effects, and when probing interactions in moderation analysis. Various measures of effect size for indirect effects are generated in mediation models, along with bootstrap confidence intervals for effect size inference. An option is available for partialing out contextual level variation when individual data are nested under a higher-level organizational structure. Individual paths in moderated mediation models can be estimated as moderated by one or two variables either additively or multiplicatively. Some models estimated by PROCESS allow up to four moderators simultaneously. See Hayes (in press) for an introduction to the use of PROCESS as well as the basic concepts and principles of mediation, moderation, and conditional process analysis.
Preparing for Use
The PROCESS.sps file should be opened as a syntax file in SPSS. Once it has been opened, execute the entire file
exactly as is
. Do not modify the code at all. Once the program is executed, the PROCESS program window can be closed. You then have access to the PROCESS
command until you quit SPSS. The PROCESS.sps file must be loaded and reexecuted each time SPSS is opened. See
the “Examples”
section below for some examples of how to set up a PROCESS command in a syntax window. Please also read the
“Model Designation and Estimation” and “Notes”
sections below for important details pertinent to execution.
Examples
(1) Moderation
process vars = protest sexism liking tenure exper/y = liking/x = protest/m = sexism /model = 1/quantile = 1/center = 1/plot = 1/jn = 1.
Estimates a simple moderation model with the effect of
protest
on
liking
moderated by
sexism
tenure
and
exper
are included in the model as covariates
.
sexism
and
protest
are mean centered prior to analysis.
Generates the conditiona
l effects (a.k.a. “simple slopes”) of
protest
on
liking
at values of
sexism
equal to the 10
th
, 25
th
, 50
th
, 75
th
, and 90
th
percentiles of the distribution in the sample.
Produces a table of estimated values of
liking
for various values of
protest
and
sexism
Implements the Johnson-Neyman technique to identify the values on the continuum of
sexism
at which point the effect of
protest
transitions between statistically significant and nonsignificant at the 0.05 level.
(2) Moderation
process vars = mathprob gender explms treat/y = mathprob/x = treat/m = explms /w = gender/mmodval = 4/model = 3.
Estimates a moderated regression model predicting
mathprob
from
treat
while including a three way interaction between
treat
,
explms
, and
gender
in the model.
Generates the conditional effect of
treat
(a.k.a. “simple slopes”) on
mathprob
for both males and females when
explms
= 4.
SPSS PROCESS documentation, 2 January 2013 Page 3
(3) Simple Mediation
process vars = cond pmi react/y = react/x = cond/m = pmi/model = 4/total = 1 /effsize = 1/boot=10000.
Estimates the total and direct effect of
cond
on
react
, as well as the indirect effect of
cond
on
react
through
pmi
.
Generates a bias-corrected 95% bootstrap confidence interval for the indirect effect using 10,000 bootstrap samples.
Produces point and bias-corrected 95% bootstrap confidence interval estimates of various indices of effect size for the indirect effects.
(4) Multiple Mediation with Mediators operating in parallel
process vars = know educ attn elab sex age/y=know/x=educ/m=attn elab/model=4 /contrast=1/normal=1/conf=90.
Estimates the direct effect of
educ
on
know
, as well as the total and specific indirect effects of
educ
on
know
through
attn
and
elab,
with
attn
and
elab
functioning as parallel mediators.
sex
and
age
are included in the model as covariates.
Produces the Sobel test for the specific indirect effects.
Generates 90% bias-corrected confidence intervals for the indirect effects using 1,000 bootstrap samples.
Calculates the difference between the two specific indirect effects and produces a bootstrap confidence interval for the difference.
(5) Multiple Mediation with Mediators operating in Serial
process vars = cond import pmi react nbhrhood/y = react/x = cond/m = import pmi /model = 6/hc3 = 1/effsize = 1/total = 1/boot = 10000/cluster = nbhrhood.
Estimates the total and direct effect of
cond
on
react
, as well as the total and all possible specific indirect effects of
cond
on
react
through
pmi
and
import.
pmi
and
import
function as mediators in serial, with
import
affecting
pmi.
Standard errors for model coefficients are based on the HC3 heteroscedasticity-consistent standard error estimator.
Generates 95% bias corrected bootstrap confidence intervals for the indirect effects using 10,000 bootstrap samples.
Produces point estimates and bias corrected 95% bootstrap confidence interval for various indices of effect size for the indirect effects.
With cases nested within neighborhoods (coded with a variable named
nbhrhood
), partials out neighborhood-level effects from all effect estimates.
(6) Moderated Mediation
process vars = frame euskept peffic risk turnout/y = turnout/x = frame/m = risk /w = euskept/z = peffic/model = 68/boot = 20000/wmodval = 2/center = 1.
Estimates the direct effect of
frame
on
turnout
, as well as the conditional indirect effects of
frame
on
turnout
through
risk.
The effect of
frame
on
risk
is modeled as multiplicatively
PROCESS is written by Andrew F. Hayes, http://www.afhayes.com Page 4
moderated by both
peffic
and
euskept
, and the effect of
risk
on
turnout
is modeled as moderated by
euskept
.
euskept
,
peffic
, and
frame
are mean centered prior to analysis.
Calculates the conditional indirect effects of
frame
on
turnou
t through
risk
among cases 2 units above the sample mean on
euskept
and with values of
peffic
at the
sample means as well as with
peffic
one standard deviation above and below the sample mean.
Generates bias corrected 95% bootstrap confidence intervals for the conditional indirect effects using 20,000 bootstrap samples.
(7) Moderated Mediation
process vars = calling livecall carcomm workmean jobsat/y = jobsat/m = carcomm workmean/x = calling/w = livecall/model = 7/boot = 5000/seed=34421.
Estimates the direct effect of
calling
on
jobsat
, as well as the conditional indirect effects of
calling
on
jobsat
through
both
carcomm
and
workmean
operating in parallel.
The effects of
calling
on both
carcomm
and
workmean
are modeled as moderated by
livecall
.
Produces the conditional indirect effects
calling
when
livecall
is equal to the sample mean as well as plus and minus one standard deviation from the mean.
Generates bias corrected 95% bootstrap confidence intervals for the conditional indirect effects using 5,000 bootstrap samples.
Seeds the random number generator for bootstrap sampling with the value 34421.
(8) M
oderated Mediation (and “Mediated Moderation”)
process vars = protest sexism respappr liking age sex/y = liking/x = protest /m = respappr/w = sexism/model = 8/boot = 5000/quantile = 1/percent=1.
Generates model coefficients for model 8, estimating the effect of
protest
on
liking
directly as well as indirectly through
respappr
, with both direct and indirect effects moderated by
sexism
.
The effect of
protest
on
respappr
is modeled as moderated by
sexism
.
age
and
sex
are included in the model as covariates.
Generates 95% percentile-based confidence intervals based on 5,000 bootstrap samples for the conditional indirect effect of
protest
at the 10
th
, 25
th
, 50
th
, 75
th
, and 90
th
percentile values of
sexism
.
Produces the indirect effect of the product of
protest
and
sexism
on
liking
through
respappr
, along with a percentile-based 95% bootstrap confidence interval.
Model Designation and Estimation
PROCESS can estimate many different models, and which model is estimated is determined by the
num
argument in the required
/MODEL
subcommand. The models that PROCESS can estimate are depicted conceptually and in the form of a path diagram in the Appendix, along with their corresponding model number as recognized by PROCESS in the
/MODEL
subcommand. Each model has certain minimum requirements as to which variables must be designated and provided in the PROCESS command. Any variable in the data set that appears in the model must be listed in the
varlist
argument of the PROCESS command (e.g.,
vars = xvar yvar mvar wvar
). Furthermore, all models require
SPSS PROCESS documentation, 2 January 2013 Page 5
a single outcome variable
yvar
listed in the
/Y
subcommand (i.e.,
/y
=
yvar)
,
where
yvar
is the name of the variable in your data functioning as
Y
in the model
a single antecedent causal agent
xvar
listed in the
/X
subcommand (i.e.,
/x = xvar)
,
where
xvar
is the name of the variable in your data functioning as
X
in the model
either a single moderator (models 1, 2, and 3) or at least one mediator (models 4 and higher) specified in the
mvlist
in the /
M
subcommand (i..e, /
m = mvlist
)
, where
mvlist
is the name of the variable or variables in your data functioning as moderator (models 1, 2, and 3) or mediator(s) (models 4 and higher). Other than
/X
,
/Y
,
/M
,
/MODEL
, and
/VARS
, the remaining required inputs to PROCESS will be model dependent. In general, any variable that is a part of the conceptual model in the model template must be provided as an input to PROCESS, and any variable that is not a part of the conceptual model must be left out unless such variables are to be treated as covariates by inclusion in
varlist
. For instance, observe that model 21 has, in addition to
X
,
M
, and
Y
, two moderators,
W
and
V
. Thus, PROCESS must also be told which two variables in the data set correspond to
W
and
V
in the diagram. This would be done with the use of the
/W
and
/V
subcommands (e.g.,
/W = wvar/V = vvar
), where
wvar
and
vvar
are the names of the variables in the data file corresponding to
W
and
V
. The
/Y
,
/X
,
/W
,
/Z
,
/V
, and
/Q
subcommands each allow only one variable, and a variable can be listed in one and only one of the
yvar
,
xvar
,
mvlist
,
wvar
,
zvar
,
vvar
, and
qvar
arguments. For instance, a variable cannot be listed as both
W
in
wvar
and
M
in
mvlist
. However, both would have to appear in
varlist
. Variable names listed in the
varlist
,
yvar
,
xvar
,
mvlist
,
wvar
,
zvar
,
vvar
, and
qvar arguments
must match the case (i.e., uppercase, lowercase, or combinations thereof) of the variabl
es in the dataset. So “ATTITUDE,” “Attitude”, and “AttiTuDe” are different variables
according to PROCESS. Thus,
/y = ATTITUDE
will produce an error even if “Attitude” exists
in your data file. In addition, the potential for errors at execution is increased when variable names are more than eight characters in length. Thus, the user is advised to reduce all long variable names in the data set and that are to be used in a PROCESS command down to eight characters at maximum. Although PROCESS has a number of error trapping routines built in, it will not catch all errors produced by improper formatting of a PROCESS command, improper listing of variables and variable names, and so forth. Any errors it has trapped will be displayed in an errors section of the PROCESS output. Errors it has not successfully trapped will appear as a long list of SPSS execution errors that will be largely unintelligible.
Multiple Mediators
All mediation models (models 4 and higher) can have up to 10 mediators operating in
parallel
, with the exception of model 6, which is restricted to between 2 and 4 and models the mediators as operating in
serial
. Mediators operating in parallel are all modeled as affected by
xvar
and, in turn, affect
yvar
, but they are not modeled to transmit their effects to any other mediators in the model. Mediators operating in serial are linked in a causal chain, with the first mediator affecting the second, the second the third, and so forth. The order of the mediators in
mvlist
is not consequential to the estimation or the model except in model 6. In model 6, the first variable

Search

Similar documents

Tags

Related Search

We Need Your Support

Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks

SAVE OUR EARTH

We need your sign to support Project to invent "SMART AND CONTROLLABLE REFLECTIVE BALLOONS" to cover the Sun and Save Our Earth.

More details...Sign Now!

We are very appreciated for your Prompt Action!

x