Hi all,

I have a multilevel logistic model (using melogit in Stata 15.1) written as:

Code:
 melogit dv i.intl_first c.totalattended c.distress || center: || ther: intl_first
intl_first is a binary variable where 1 = international students and 0 = students from US. I am trying to get the random slope value for both groups (i.e., 1 and 0). I tried this code in an attempt to get the random components by group:

Code:
 xi: melogit dv i.intl_first c.totalattended c.distress || center: || ther: i.intl_first
That, however, produces the same values as the original command code. That is, the random slope for ther: intl_first is .15 in both models.

I tried changing the intl_first coding so that 1 = student from US and 0 = international student, which produced a different random slope in the output table. Would that represent the random slope for those in the student from US group? If not, is there an effective way to write the command in order to obtain the random slope by group?

I have also used:

Code:
 predict re*, reffects
to obtain the BLUPs for the random components of the model. Is it appropriate or meaningful to compare the BLUPs by the binary grouping variable (intl_first)?