Hi all,

I've got a dataset of students in schools for which I've imputed 20 datasets, and I'm estimating the effect of an individual attribute x2 on a response variable y along with a cross-level interaction effect of that individual attribute with a binary level 2 variable x3 that distinguishes between two types of schools. I've fit a two-level mixed effects model of response variable y on several individual level predictors and a random intercept at the school level. It's essentially the same as xtreg, re.

Code:
mi estimate: mixed y  x1 x2 i.x3  c.x2##i.x3  || school : , mle vce(cluster school)
I want to plot the effect of x2 on y by each type of school, and I'd like the predicted values to include the random intercept u_i. Although I can calculate the marginal effects for the continuous variable x2 interacted with the binary variable x3 using:

Code:
mimrgns i.x3, at(x2=(-5(0.1)5)) atmeans asbalanced cmdmargins
marginsplot
While mimrgns allows me to use marginsplot in the MI context, like margins, it does not allow for including u_i.

Is there any way, in the MI context, to do this?

Thank you,
Kerby