How do I add covariates to a SEM model

[Al Bothwell]

I have data from an experiment on physical activity. Participants were assigned to one of two groups (control or intervention). There are five mediators each measured using a likert-type scale. The outcome is a continuous measure. I used sembuilder and this is the syntax:

sem (Mediator1 -> DV, ) (Mediator2 -> DV, ) (Mediator3 -> DV, ) (Mediator4 -> DV, ) (Mediator5 -> DV, ) (IV -> DV, ) (IV -> MV1, ) (IV -> MV2, ) (IV -> MV3, ) (IV -> MV4, ) (IV -> M5, ), nocapslatent

How do I add five covariates (CV1 CV2 CV3 CV4 CV5 CV6 ) to the model? I will prefer to do this in sembuilder but syntax also works well.

I really need help with this.

Al Bothwell

[wbuchanan]

So it doesn't appear that you're fitting a mediation model. It isn't clear whether MV1 or Mediator1 is your actual mediator, and it's much safer to use lower cased variable naming conventions to avoid the possibility that the variable is interpreted as a latent variable. Assuming mv# would represent your mediators your model would look something more like: sem (dv mv1 mv2 mv3 mv4 mv5), nocapslatent Now you would have causal paths from the IV to each of the mediators and DV as well as paths from each of the mediators to the DV. (I'm doing this from my phone so you'd likely want to test this first). Then you can use the same syntactic structure to specify the equations representing the relationship(s) between your covariates and their respective dependent variables. But it isn't clear from your post exactly how/what the relationship of the covariates is. I would also consider thinking of your mediator as a single latent construct or consider using gsem in order to estimate the IV to MV equations using the correct estimators (e.g., if you're using some type of "Likert" scale the values are more than likely ordinal in nature).

[Al Bothwell]

Thank you!