Message/Author |
|
|
Hello again, If I create an interaction term with my exogenous variables, for example, define x1xgen = x1*gender. and then regress my (latent) endogenous variables on the exogenous variables as such: Factor1 ON x1 gender x1xgen; do I allow the interaction terms to covary or do I set the relationship to zero, i.e., x1 WITH x1xgen@0; Thank you for this service. |
|
|
The interaction terms covary without you saying anything because they are "x variables". |
|
|
Yes, that makes sense. But this is ok in the context of my model, to have the x variables covary? I prefer not to constrain them but I am not used to seeing this in regression w/ interactions (as I am relatively new to Mplus/SEM). Thank you. |
|
|
I don't think it is ok - why would you want x1 to have zero correlation with x1*gender? |
|
|
I agree. Thank you. A follow-up question is that the SR model contains method factors as there are multiple informants (child, parent) as well as random intercept factors. I am tempted to just let my x variables covary with these. Is there any reason why I should constrain these relationship? Thanks! |
|
|
No, just covary or regress the factors on the covariates. |
|
|
Thank you. So I should constrain the correlation between x variables and these factors to O? |
|
|
No - the opposite. |
|
|
ok, thank you. |
|
Back to top |