Hi all,
I'm running a path analysis in AMOS which contains 2 interactions on two paths coming from the same DV. For each interaction, the 2 variables and the product of the zscores of these two are entered and covaried. I've attached a representation of the sort of model I'm trying to run in word (with covariances included in grey).
My question is whether the interactions and the variables should always be covaried with each other when on the same level of the path analysis?
Many thanks,
Kathryn
Kathryn,
yes, all predictor variables should be allowed to correlate. Packages such as Mplus include these parameters by default. You need to do this because you'll want to see if the interactions explain variance in the DVs above and beyond the 'main effects' of the predictors.
Mark
Thank you Mark, Young, and Fen for your answers. I will now allow all predictor variables to correlate. A follow up question for this - my variables are observed so AMOS only allows covariance of errors in the mediators and level 1, should I covary predictor variables with this error/use this error for any interaction of the effect of the mediator on outcome?
Thanks again all!
- Badges
- Science topic
- Similar topics
- Instrumentation Engineering
- Instruments