Am I correct that, at present, there is no way to compute an interaction term with a latent variable in the DSEM framework (due to the inability to use the XWITH command)?
I am testing a cross-lagged two-level model, in which the two primary variables are observed and binary. I have been modeling these as single indicator latent variables in order to use the DSEM framework, because, as I understand it, DSEM does not work for categorical outcomes at this time.
Based on this, it seems that it's not possible for me to test moderation at either the within- or between- level - as doing so would require the use of the XWITH command to create a product of the moderator and the latent variables, but that command is not available yet in DSEM.
Am I correct in my interpretation? If not, I would love to know how to test the impact of a dimensional/interval, within-person (level 1) moderator on two latent variables, also at the within-person level.
XWITH is not available but some moderation models are available when using random slopes predicted by another variable. XWITH can also be pursued through the use of plausible values with two step estimation http://statmodel.com/download/Plausible.pdf