|
|
| Message/Author |
|
|
| Stanislaw posted on Monday, September 02, 2019 - 3:17 am
|
|
|
Hallo,
I am testing a mediation model with two predictors (one is intervention vs. control group), a mediating variable, and an outcome variable. Below you can find the syntax. Do I need to consider clusters in the mediating and outcome variable (and thus, use type=complex)? The existing clusters are due to the intervention, I guess.
analysis: type = general;
estimator=mlr;
model: quan on instr zsw;
mod on quan instr;
[zsw];
[instr];
model indirect: mod ind zsw;
mod ind instr;
output: sampstat stdyx;
Thank you for your help! |
|
|
|
|
| What are your clusters? |
|
|
| Stanislaw posted on Tuesday, September 03, 2019 - 12:05 am
|
|
|
| Sorry: Students in classes. One half of each class got strategy instructions, the other half not. We have design effects greater than two for strategy use (mediator)-of course. |
|
|
|
|
You can do twolevel modeling in line with Preacher et al's pdf's on our Mediation page:
http://www.statmodel.com/Mediation.shtml
They call your case 1-1-1 mediation (X, M, and Y are all on level 1).
You should also read the articles that go with those pdf's. |
|
|
| Back to top |
|
|
