Moderation with latent variables in Stata 13 I am wondering if anyone has insight regarding the creation of interaction terms (moderators) in a model with latent predictors using Stata 13 - SEM. I have what I want to be the moderator set up currently as a mediator and need to test competing models. I am not sure if I need to run a multiple group model or if there is a cleaner approach. Any thoughts would be helpful. 
 A: You shouldn't use the multigroup approach if the moderating variable is continuous. You can do it manually using one of several methods that are described in the literature. The most popular are the so-called constrained approach, and two variants of the unconstrained approach: mean-centered and residual-centered. Of these the easiest to implement is the mean-centered unconstrained approach, in particular the double-mean centered variant. Here are three references you might find helpful:
Marsh, H. W., Wen, Z., & Hau, K.-T. (2004). Structural equation models of latent interactions. Psychological Methods, 9(3): 275–300.
Marsh, H. W., Wen, Z., Hau, K.-T., Little, T. D., Bovaird, J. A., & Widaman, K. F. (2007). Unconstrained structural equation models of latent interactions: Contrasting residual- and mean-centered approaches. Structural Equation Modeling, 14(4): 570–580.
Lin, G.-C., Wen, Z., Marsh, H. W., & Lin, H.-S. (2010). Structural equation models of latent interactions: Clarification of orthogonalizing and double-mean-centering strategies. Structural Equation Modeling: A Multidisciplinary Journal, 17(3): 374–391.
