Hello, I'm trying to build a mixed-effects model with the following vars: Fixed effects - race, gender, age group, region, orthopaedic component, season, comorbidity. Random effects - year of procedure, region, hospital npi, surgeon. Outcome - 90-day readmission. I'm not sure if I have the proper syntax to properly specify random effects. There is nesting (i.e. random intercept / subject = surg(npi region yr)), but each of these variables are also random effects. For the model I wish to build, does anyone have any suggestions on how they might code/specify random effects? Code proc glimmix data=in method=laplace; class race (ref="1"); class gender (ref="1"); class Age_Group (ref="1"); class Region (ref="1"); class Component_Major (ref="1"); class Season (ref="1"); class Comorbidity (ref="1"); class yr; class surg; class npi; class region; model Read90 (event='1') = race gender Age_Group region component_major season Comorbidity / dist=binary link=logit solution oddsratio; *yr, region, npi, surg are all random effects; *surgeons clustered within hospitals clustered within regions clustered within years; random intercept / subject = yr; random intercept / subject = region; random intercept / subject = npi; random intercept / subject = surg; random intercept / subject = surg(npi region yr); lsmeans racekey gender Age_Group region component_major season Comorbidity / oddsratio ilink diff cl adjust=bon; run; Thank you!!
... View more
