| Message/Author |
|
|
| Anna Bayne posted on Friday, January 06, 2017 - 6:12 am
|
|
|
Dear Dr. Muthen,
I am running a multilevel sem with two IVs and three mediators. I have three DVs but analyze each seperately. I would like to assess if one of the three mediators is more important than the other two. Is there any way to assess whether one mediator is more important than the other two? Or also just whether one is more important than one other? Your help is much appreciated.
My syntax looks as follows:
VARIABLE:
USEVARIABLES = id x1 x2 m1 m2 m3 y;
MISSING ARE ALL (-9);
CLUSTER IS id;
DEFINE:
CENTER x1 x2 m1 m2 m3(GRANDMEAN);
ANALYSIS:
TYPE IS TWOLEVEL;
MODEL:
%WITHIN%
dv ON x1 (c1);
m1 ON x1 (a1);
dv ON x2 (c1b);
m1 ON x2 (a1b);
dv ON m1 (b1);
m2 ON x1 (a2);
m2 ON x2 (a2b);
dv ON m2 (b2);
m3 ON x1 (a3);
m3 ON x2 (a3b);
dv ON m3 (b3);
x1 WITH x2;
m1 WITH m2;
m1 WITH m3;
m2 WITH m3;
MODEL CONSTRAINT:
NEW(ind1 ind2 ind3 ind1b ind2b ind3b);
ind1=a1*b1;
ind2=a2*b2;
ind3=a3*b3;
ind1b=a1b*b1;
ind2b=a2b*b2;
ind3b=a3b*b3;
OUTPUT:
stand (STDYX);
CINTERVAL; |
|
|
|
|
Note that you have no Between part to your twolevel model. Perhaps you want to use Type=Complex instead.
Mplus does not allow testing if an indirect effect is more important than another. You can test if they are equal but that assumes the same scale for the variables involved or comparing standardized estimates.
You may want to discuss this on SEMNET. |
|
|
| Anna Bayne posted on Monday, January 09, 2017 - 1:31 am
|
|
|
Dear Dr. Muthen,
thank you very much for your reply and advise, I will take it into account.
Can you tell me what I would need to add to the syntax in order to test if the indirect effects are equal? I would compare the standardized estimates then.
I highly appreciate your suggestions.
Kind regards,
Anna Bayne |
|
|
|
|
You can for instance add a NEW parameter diff12 in Model Constraint:
diff12 = ind1 - ind2;
If you want to compare standardized versions you would have to express the relevant variances and do the standardization in Model Constraint. |
|
|
| Back to top |
