In general, patients don't change cetert in a multicenter follow up study, we can set 'random intercept / subject=center; random intercept /subject=patient(ceter)' to account center variability. What if some patients changed center in middle of follow up (a patient had test in center1 in visit 1 and 2 then mobed to center2 for visit 3 and 4)? proc glimmix; class patient center; model outcome=visit; random intercept / subject=center; random intercept /subject=patient(center); run; Thanks, HL
... View more