|
|
 |
| Graphical presentation of interaction... |
 
|
| Message/Author |
|
|
|
|
Dear Professor Muthen,
We would like to present significant interaction effects graphically. In fact, the reviewers ask us to plot the interaction effects and report simple slope coefficients. In our models dependent variables are continuous and both independent and moderating variables dichotomous. Is it possible to draw the figures using Mplus (8.0) and how can we do it?
Thank you very much for your help. |
|
|
|
|
See our Mediation page at
http://www.statmodel.com/Mediation.shtml |
|
|
|
|
Thank you for this model. I've tried to apply it to our moderating model:
..
USEVARIABLES: sex age edu int padT1 padT2, intedu ;
MISSING ARE ALL (-99);
CLUSTER IS org;
DEFINE: intedu = int*edu ;
ANALYSIS:
TYPE = COMPLEX;
ESTIMATOR=MLR;
!edu is a dichotomous moderator with values 0=low, 1=high.
MODEL:
padT3 ON sex age padT1 edu int intedu;
MODEL CONSTRAINT:
LOOP (edu, 0,1,0.1);
PLOT: TYPE=PLOT2;
OUTPUT: STAND;
There is something wrong with the LOOP command (at least), but I don't know what.
Thank you again. |
|
|
|
|
Take a look at the Mediation page again - note that the type of input is
MODEL CONSTRAINT:
LOOP(mod,-2,2,0.1);
PLOT(indirect);
indirect = b*(gamma1+gamma2*mod);
- you don't have the PLOT part. Also, because edu is 0/1 you should have LOOP(edu, 0, 1, 1) because you step from 0 to 1. |
|
|
|
|
Thank you for the LOOP instructions.
However, I don't understand how can I use the indirect part of the model, as we don't have any mediator in our model. I thought that the "m" in the model you referred is the mediator...?
Our model without mediator is:
padT3 ON sex age padT1 edu int intedu;
Y=padT3
X=padT1, sex, age, int
Z=(moderator) edu
XZ=(interaction term) intedu
!sex, age, int and edu are dichotomous variables.
Thank you for your patience. |
|
|
|
|
Just drop the "b" part and call it something else than "indirect" and you have it.
You can also see the Chapter 1 regression example inputs from our RMA book on our website at
http://www.statmodel.com/Mplus_Book_Tables.shtml |
|
|
|
|
If I am correct, our model would be:
padT3 ON int (gamma1)
intedu (gamma2)
edu
sex age padT1 ;
MODEL CONSTRAINT:
LOOP(edu,0,1,1) ;
PLOT(TTT)
TTR=gamma+gamma2*edu ;
PLOT: TYPE=PLOT2
!Y=padT3
!X=padT1, sex, age, int (intervention group, as this is a RCT study)
!Z=(moderator) edu
!XZ=(interaction term) intedu
The model estimated normally and I am satisfied with the model.
However, I don't know how to read the figure. X axis is the moderator (edu, values vary between 0 to 1), but what do the Y axis (TTT) and the three separate lines describe? [Usually the Y-axis would be the dependent variable (in this case PADT3), X-axis the intervention (int 0=control group, 1=intervention group) and the moderating effect would be presented as separate lines (0=low edu, 1=high edu). I can change the places of int and edu in this model, so that the X-axis is then the intervention, but still the figure is difficult to interpret.
Best regards,
Mervi |
|
|
|
|
First, note that you are not plotting Y when you plot TTT - you are plotting only the simple slope. For instance, your TTT expression does not contain the Y intercept. You may want to read Chapter 1 of our RMA book.
Regarding the 3 lines, we need to see in detail what you are doing - send your output and gh5 files to Support along with your license number. |
|
|
| Back to top |
|
|
