| Message/Author |
|
|
|
|
I am struggling with my moderation model. All my variables are at the within-level and continuous. I would like to test if women's shame (ShameW) moderates the associations between women's empathy (EmpW) and partners' empathy (EmpP) on women's pain (Pain) and both partners' distress (DetW and DetP) :
USEVARIABLES = EmpW EmpP Pain DetW DetP ShameW IntW IntP;
CLUSTER = Couple ;
WITHIN
EmpW EmpW ShameW IntW IntP Pain DetW DetP ;
DEFINE:
IntW=EmpW*ShameW ;
IntP=EmpP*ShameW ;
CENTER
EmpW EmpP ShameW Pain DetP DetP
(GROUPMEAN);
ANALYSIS:
TYPE = twolevel random ;
MODEL:
%BETWEEN%
%WITHIN%
Pain ON EmpW ShameW EmpP IntW IntP;
DetW ON EmpW ShameW EmpP IntW IntP;
DetP ON EmpP ShameW EmpA IntW IntP;
Pain WITH DetW DetP ;
DetP WITH DetW ;
EmpW WITH EmpP ;
MODEL CONSTRAINT:
NEW (LOW_M MED_M HIGH_M);
LOW_M = 0.85; ! - SD
MED_M = 3.37; ! mean value
HIGH_M = 5.89; ! + SD
I am stuck here. I probably have to calculate simple slopes for each value of the moderator and use loop plot to plot model for low, med, high values of the moderator, but I don't know how to write the syntax so I could have the interaction effects.
If this syntax is not the best, do you have another one? |
|
|
|
|
You can use syntax in line with the RMA book examples for chapter 1 at
http://www.statmodel.com/mplusbook/chapter1.shtml
See the script for:
Table 1.8, Part 2, p. 31: Regression with a randomized intervention using treatment-baseline interaction and LOOP PLOT of moderation |
|
|
|
|
See also the first example of the Topic 11 Short Course video and handout on our website at
http://www.statmodel.com/course_materials.shtml |
|
|
|
|
Thank you very much for such a quick response!
I looked at the script in Table 1.8, part 2, p. 31. I fear that what I am struggling with is the formula, which directly translate into my syntax.
I understand with a dyad dataset (no moderator), we would have two equations in SEM:
Y1 = b0 + b1X1 + b2X1 + e1
Y2 = b0 + b2X1 + b1X1 + e2
and with a moderator we would have (correct me if I am wrong) :
Y1 = b0 + b1X1 + b2X2 + b3M + b4X1M + b5X2M + e1
Y2 = b6 + b7X2 + b8X1 + b9M + b10X2M + b11X1M + e2
In your example you wrote :
model constraint:
LOOP(x,-1,1,0.1);
PLOT(effect);
effect = b1 + b3*x;
However, could you help me with the 'effect formula' part? Should I have two lines here (dyad dataset)?
Lastly, you set 0.1 increment in your example. Is it the usual choice? |
|
|
|
|
Your 2 moderator equations look correct. The moderator functions are:
Y1: (b1+b4*M) for X1 and (b2+b5*M) for X2
Y2: (b8+b11*M) for X1 and (b7+b10*M) for X2. |
|
|
|
|
Does this looks correct? I created equations for both y1 and y2 DVs. Only y1 are shown due to size limit.
MODEL CONSTRAINT:
NEW (LOW_M MED_M HIGH_M);
NEW (SIMP_LOy1 SIMP_MEDy1 SIMP_HIy1 SIMP_LOy2 SIMP_MEDy2 SIMP_HIy2);
LOW_M = 0.85;
MED_M = 3.37;
HIGH_M = 5.89;
SIMP_LOy1 = (b1 + b4*LOW_M)+(b2 + b5*LOW_M);
SIMP_MEDy1 = (b1 + b4*MED_M)+(b2 + b5*MED_M);
SIMP_HIy1 = (b1 + b4*HIGH_M)+(b2 + b5*HIGH_M);
! Same but with y2 equations.
PLOT(LOMODy1 MEDMODy1 HIMODy1 LOMODy2 MEDMODy2 HIMODy2);
LOOP(ShameW,1,5,0.1);
LOMODy1 = (b0 + b3*LOW_M) + ((b1 + b4*LOW_M)+(b2 + b5*LOW_M))*ShameW;
MEDMODy1 = (b0 + b3*MED_M) + ((b1 + b4*MED_M)+(b2 + b5*MED_M))*ShameW;
HIMODy1 = (b0 + b3*HIGH_M) + ((b1 + b4*HIGH_M)+(b2 + b5*HIGH_M))*ShameW;
!Same but with y2 equations. |
|
|
|
|
No, you should not add the 2 terms (b1 + b4*LOW_M)+(b2 + b5*LOW_M); they represent different things - the first part is the X1 effect and the second part is the X2 effect. You can look at each separately.
I think you should consult with a statistician - someone who is familiar with moderation modeling. We can not teach it here. |
|
|
| Back to top |
