I've got a mixed effects model with both crossed and nested random effects. The data concern a series of multi-item tests taken by the participants. Viewed as a spreadsheet, each observation (Level 1) is nested horizontally in participant (Level 2H) and simultaneously nested vertically in test item (Level 2V) which is in turn nested in test (Level 3V).

To handle both the crossed and nested effects, I can just use the "_all:R.factor" trick to make my model:

Code:
mixed outcome predictor || _all:R.participant || test: || item:
If I had a continuous outcome variable, I could then do what Isabel Canette describes in this blog post to get a random slope for each participant:

Code:
tab ID, gen(id_ID)  
unab idvar: id_ID*  
foreach v of local idvar {    
   gen predictor_`v' = predictor*`v'    
   }  

mixed outcome predictor || _all:R.ID || _all: predictor_*, cov(identity) nocons || test: || item:
All of that seems to work with MIXED and a continuous outcome, but the problem is that I have a categorical outcome (either binary or ordinal).

When I try the same commands but substitute MELOGIT for MIXED, the model just gives me identical coefficients for all the participants.
Is there a way to do what Canette is describing but with MELOGIT or MEOLOGIT?