I have a survey over 14 schools. In each school, 9 to 11 students were interviewed for a comparison of two items A and B, which one they prefer. The outcome Y is a binary variable with 1 if a student prefer item A (compared with B) and 0 otherwise. Four variables have been measured: "School_Status" and "School_size" at school-level and "Age" and "English_proficiency" at student-level. School_Status has 3 levels, School_size 3 levels, Age 2 levels and English_proficiency 2 levels. The total number of observation was 155.

I used the following SAS glimmix code:

Proc glimmix;    class School_Status School_size Age English_proficiency;    model Y (event='1')= School_Status School_size Age English_proficiency / s dist=binary;    random intercept  Age English_proficiency / type=un sub=school;run ;

It did not converge.

When I used instead:

Proc glimmix method=laplace;

I received the following message:

ERROR: QUANEW Optimization cannot be completed

Grateful if someone could help.