For my learning, I tried the code regarding “If you used the code: estimate 's2' | int 1 / subject 0 1 cl ilink; for the situation with fixed effects (you give a categorical and a continuous predictor in your code), then you are getting the EBLUP for the last level of the categorical variable (always a coefficient of 0) and for the continuous predictor with a value of 0” I wonder if you mean estimate 's2' int 1 | int 1 / subject 0 1 cl ilink; is equivalent to: estimate 's2' int 1 school_type 0 1 entry_score 0 | int 1 / subject 0 1 cl ilink; If I run the following code using the dataset I sent, different results show. I was also removing 'int 1' before vertical bar (this is in your statement) but I could not get the same results from the two estimate statements. PROC glimmix data=blup.school NOCLPRINT MAXLMMUPDATE=100 ; class school school_type; model final_pass = entry_score school_type / s cl dist=bin link=logit ; random intercept/ subject=school solution cl ; estimate 's2: statement 1' int 1 | int 1 / subject 0 1 cl ilink; estimate 's2: statement 2' int 1 school_type 0 1 entry_score 0 | int 1 / subject 0 1 cl ilink; run; Label Estimate Standard Error s2: statement 1 0.01563 0.8291 s2: statement 2 -0.7348 0.8021
... View more