David A. Kenny
June 29, 2014
APIM with Distinguishable Dyads Macro
This
moderation macro, called APIMDtext, was written by David A. Kenny, Department
of Psychology, University
of Connecticut
. This is
Version 1, completed on May 25, 2010 and revised on May 15, 2011 and May 22, 2012 and again on January 18, 2013. This
marco may well contain errors and will certainly be revised, and so it is
advisable to return for updates.
A correction was made on June 29, 2014. This macro has been replaced by shiny app that can be accessed at APIM_MM and an extensive description is available on here.
The
APIM is a model for dyadic data. It is
assumed that the data have a pairwise structure. The dataset needs to be in pairwise format.
To convert an individual file to a pairwise data click here. One variable has an effect on another and
that effect can be either the effect from one’s own score, the actor effect, or the effect
from one’s partner, the partner effect. It
is assumed that the dyad members are distinguishable, e.g., one member is the
husband and the other member is the wife.
Thank You!
I
thank my good friend Linda Acitelli for the sample data. I also than Jessica Fales who previously pointed out an error to me.
Download:
APIMDText.SPS (You need SPSS to open this and the next
3 files.)
Sample
Data File
(Best to give plural labels to the two levels of the distinguishing variable, e.g., "Husbands" and "Wives".)
Macro Call
SPSS
Output
Macro Output (You do not need SPSS to open this
file. But if you do not use “wordwrap” it will look ugly.)
To understand how
to run a macro return to the DataToText page.
The macro takes a few minutes to run and so be patient. Make sure to
backup the raw data file, as sometimes an error in the macro can alter the data
file.
PLEASE
READ THIS: Note that the
output (the text file you name) cannot be viewed in SPSS and is not written to
the SPSS output file. You need to view it in text reader (e.g.,
Notepad). When viewing the file, make sure to use “wordwrap” and Courier
or New Courier font.
The Macro Call
This is the
statement for the sample data:
APIMDText a = RSpouse/p = PSpouse /y =
RSatisfied /distvar=Rgender /dyadid = coupleid
directory='c:\' xn = 'Other Positivity' yn = 'Satisfaction' clist=
yearsmar.
The defaults are
as follows:
a = Actor
p = Partner
y = Outcome
distvar = DistVar
directory ='C:\'
xn ='X'
yn ='Y'
alpha =.05
ofile ='APIMDtext.txt'
clist =
That is, if you
just say “APIMDText.”, the program will assume that are variables in the SPSS
data file with variables named Actor, Partner, Outcome, and DistVar.
The data file
should be open; values should be given for both levels of the distinguishing
variable in the SPSS file.
APIMDtext was
written on SPSS 16 and 18 and there is no guarantee that it work on earlier or
later versions of SPSS. It appears that the three tables at the end of the file
are not correct if SPSS 15 or earlier is run.
The macro:
allows for only a single actor and partner
variable, distinguishing variable, and outcome,
presumes that effects are linear and not
non-linear (e.g., quadratic)
presumes that the outcome variable is
measured on an interval scale, and
uses listwise deletion.
Variables in the
macro:
a = name of the causal variable
for the person in the SPSS data set (do not use quotes for any of the SPSS
names)
p = name of the causal variable
for the partner in the SPSS data set
y = name of the moderating
variable in the SPSS data set
distvar
= name of the distinguishing variable; must be a dichotomy and each member must
have different scores; labels for the two categories should be provided
directory = the name of the
directory where temporary files are written (use quotes); this must be a
directory you are allowed to write on; APIMDtext will leave some files on this
directory when it is done; I am working on finding a way to erase them.
xn = name of the causal variable
in the text output (use quotes; spaces are allowed)
yn = name of the outcome
variable in the text output (use quotes; spaces are allowed)
(For the above two
variables, it is advised to use English names and not SPSS acronyms. Also
it is advised to capitalize the first letter of each word.)
alpha = significance level
(defaults to .05)
ofile = the name of the output
file (use quotes); this is where you go to find the text
clist = the SPSS names of the
covariates separated by spaces
It is safest to
give arguments for all macro variables. Note carefully what terms have
quotes and what do not and where the slashes are where they are not.
There is no
guarantee for accuracy. Examine not only DataToText output file, but also
the SPSS output file. The user needs to edit the APIMDtext output in research
reports. Please cite APIMDtext if you do use it. Please cite this ApimText webpage if you do use it. Moreover, you need
a footnote that says: “Some of the material here was produced by the SPSS macro
ApimDText (Kenny, 2011). Any material in your paper that is verbatin must be enclosed in quotes.”
If a non-English
version of SPSS is being used, APIMDtext changes the language to English.
It does not currently change the language back to the original language.
Warnings
APIMDtext
provides several possible warnings. The user needs to pay careful
attention to them.
1. With covariates, ApimDText can fail to remove
missing cases on the covariates. The
researcher should remove those cases before undertaking the analysis.
2. The outcome variable is a dichotomy and
logistic regression and not ordinary regression should be used.
3. The actor and partner variables are high
correlated and this colinearity compromises the analysis.
4. Because zero is not a possible value for the cause
variable, grand-mean centering that variable should be considered.
5. Because the causal variable is a dichotomy,
the product term and discrepancy score are perfectly correlated and only one of
the two should be reported.
Links
DataToText Page
Frequently
Asked Questions
Macro Output
If Notepad is
used make sure you use the wordwrap option in formal. Also for the tables
to align use Courier font.
The output using
sample data:
WARNINGS: 1. With covariates, ApimDText can fail to remove missing cases on the covariates.
The researcher should remove those cases before undertaking the analysis. 2. Because zero is
not a possible value for Other Positivity, grand-mean centering that variable should be
considered.
Actor-Partner Interdependence Model for Wives and Husbands.
The focus of this study is the investigation of the effect of Other Positivity on
Satisfaction and how that effect differs for Wives and Husbands. Both the effect of own Other
Positivity (actor) and the effect of partner's Other Positivity (partner) on Satisfaction for
Wives and Husbands are studied. There are a total of 148 dyads with no missing data. The
total number of individuals is 296. The means and standard deviations for Wives and Husbands
are presented in Table 1. There is one covariate that is controlled in all analyses. The
covariate explains a statistically significant amount of variance of Satisfaction controlling
for actor and partner effects (.006 proportion of the total variance for the Wives and .016
proportion for the Husbands), chi square test with 1 degree of freedom equal to 9.849 (p =
.002).
RESULTS
Actor Effects
The actor effect for Wives is equal to .384 and is statistically significant (p < .001),
with a medium effect size (beta = .386), and the actor effect for Husbands is equal to .442 and
is statistically significant (p < .001), with a medium effect size (beta = .443). (See Table 2
for the actor effect estimates.) The difference between these two actor effects is not
statistically significant (p = .584).
Partner Effects
The partner effect from Husbands to Wives is equal to .339 and is statistically
significant (p < .001), with a small effect size (beta = .323). The partner effect from Wives
to Husbands is equal to .268 and is statistically significant (p < .001), with a small effect
size (beta = .269). (See Table 2 for the partner effect estimates.) The difference between
these two partner effects is not statistically significant (p = .508).
Actor-Partner Interactions
The effect of the product of actor and partner variables on Satisfaction for Wives is equal to -.150 and is not statistically significant (p = .367). The partner effect for persons who are one standard deviation above the mean on Other Positivity is .258 and for persons who are one standard deviation below the mean on Other Positivity is .414. Additionally, the effect of the product of actor and partner variables on Satisfaction for Husbands is equal to -.356 and is statistically significant (p = .010). The partner effect for persons who are one standard deviation above the mean on Other Positivity is .101 and for persons who are one standard deviation below the mean on Other Positivity is .438. The difference between the interaction effects for Wives and Husbands is not statistically significant (p = .179).
The effect of the absolute difference of the two members' scores for the variable Other
Positivity on Satisfaction of Wives is equal to -.040 and is not statistically significant (p =
.711). Thus, if two members have the same score on Other Positivity, the score on Satisfaction
for Wives is .040 units higher than it is for a dyad whose scores on Satisfaction differ by one
unit. The effect of the absolute difference of the two members' scores for the variable Other
Positivity on Satisfaction of Husbands is equal to .022 and is not statistically significant (p
= .802). Thus, if two members have the same score on Other Positivity, their score on
Satisfaction for Husbands is .022 units lower than it is for a dyad whose scores on
Satisfaction differ by one unit. The difference between these two discrepancy effects is not
statistically significant (p = .540).
Effect of the Distinguishing Variable
The predicted score on Satisfaction for those who score zero on Other Positivity is .510
and .590 for Husbands, and that difference is not statistically significant (p = .840), with a
small effect size (d = -.190).
Relation of Actor and Partner Effects: Testing for Patterns
An analysis was made of the relative size of actor and partner effects to determine if
there were any patterns in the effects. For Wives, there is evidence for "couple model" (Kenny
& Cook, 1999) in that the actor and partner effects are not statistically significantly
different. It may make sense to sum or average the two Other Positivity scores for Wives. For
Husbands, there is evidence for "couple model" (Kenny & Cook, 1999) in that the actor and
partner effects are not statistically significantly different. It may make sense to sum or
average the two Other Positivity scores for Husbands.
Error Variances and Correlations
The correlation between Wives errors with Husbands errors is equal to .470. Thus, the
two members of the dyad are similar to one another. The error variance for Wives is equal to
.206 and for Husbands is .141. The R squared (Kenny, Kashy, & Cook, 2006), controlling for the
covariate, for the Wives is equal to .269 and for the Husbands is equal to .343. Finally, the
correlation between Other Positivity for Wives and Husbands is equal to .224.
Test of Distinguishability
The test of distinguishability yields a chi square test with four degrees of freedom that
equals 7.786 (p = .100). Because the test of distinguishability is not statistically
significant, we conclude that members are statistically indistinguishable. The test of the
effect of the distinguishing variable is not statistically significant (p = .840). The test of
the interaction of the distinguishing variable with the actor effect is not statistically
significant (p = .584), and the test interaction of the distinguishing variable with the
partner effect is not statistically significant (p = .508). Finally, the test that error
variances are different is statistically significant (p = .009).
Treating Dyad Members as Indistinguishable
In the analyses that follow, we ignore differences between Wives and Husbands. The
overall actor effect is equal to .411 and is statistically significant (p < .001), with a
medium effect size (beta = .413). The overall partner effect is equal to .299 and is
statistically significant (p < .001), with a small effect size (beta = .300). The intraclass
correlation treating dyad members as indistinguishable is equal to .465 and the R squared is
equal to .141. Treating the dyad members as indistinguishable, there is evidence for "couple
model" (Kenny & Cook, 1999) in that the actor and partner effects are not statistically
significantly different. It may make sense to sum or average the two Other Positivity scores.
The actor-partner interaction is equal to -.256 and is not statistically significant (p =
.052). The partner effect for persons who are one standard deviation above the mean on Other
Positivity is .172 and for persons who are one standard deviation below the mean on Other
Positivity is .427. Alternatively, the effect of the absolute difference of the two members on
Other Positivity is equal to -.003 and is not statistically significant (p = .975). Thus, if
two members have the same score on Other Positivity, their score on Satisfaction is .003 units
higher than it is for a dyad whose scores on Satisfaction differ by one unit. Treating dyad
members as indistinguishable, there is not evidence of an actor-partner interaction.
(Tables, figure, and references not included here.)
Return
to the Top of the Page
Return to the
DataToText Page