Hello all,

I'm examining the association between race of the defendent (5 groups) and the legal process, in different states (thus Multilevel). For this purpose, I'm running a multilevel analysis using -mixed- coomand, with random slope for the race variable.
This is what I get for the random effect parameters:

Array

As you can see, the conf. interval for the race-slope variance is not estimable.

Is it possible that this is happening due to the fact that not every race is represented in every state?
For example, some states don't include Hispanic defendents, while others don't include Asian defendents.
I checked, and if I collapse the 5 groups into 2 groups (minority v.s. majority), so every state include observations from both groups, this problem doesn't occur.
Any help would be appreciated.

Thanks,
Eran