

SEM with moderation and mediation 

Message/Author 

stella posted on Thursday, August 01, 2013  12:07 pm



Hi Dr(s) Muthens, We are interested in running an SEM model that includes both moderation and mediation. The model is as follows: A >B >C and D > E with F moderating the path between A and B; C and D as mediators of the relationship between B and E (outcomes). We also have some baseline variables (e.g., gender, ethnicity) predicting A. Originally we were considering a moderated mediation approach; however, conceptually our moderator is not moderating the relationship between B and E through C/D. We were wondering if we can test both of these in the same model, if so what would the syntax entail? We ran mediation syntax, but were wondering about how to combine this syntax with moderation syntax (i.e., how do we test moderation). Thank you in advance! 


An easy way to test moderation is if the moderator is binary (or nominal), in which case you can do a multiplegroup analysis where group is moderator categories. With a continuous moderator you would create product variables. So if F moderates the path between A and B, you would created F*A in DEFINE and then say B ON A FA. See also McKinnon's book. Not sure what you means by "test both of these in the same model". 

stella posted on Thursday, August 01, 2013  2:03 pm



Thanks for your prompt response! To clarify  "test both of these in the same model" we meant testing the moderation and mediation together in one model when the moderator occurs in the first portion of our model and the mediation is the second portion (i.e., it's not moderating the indirect effects). We realize that this might be difficult to conceptualize without visual aid  is it possible to email you a visual model? 


Visual aid not needed. First try doing it multiplegroup. 

Back to top 

