a*******g
发帖数: 80 | 1 I have a group of patients, 2 trt arms and two centers. the outcome is a
continuous variable. I am interested the effect of trt but like to control
the effect of centers. Here is the code I used:
proc glm data=test;
class recon treatment_arm;
model scores=treatment_arm|recon/ss3;
estimate "trt diff"
treatment_arm 1 -1
treatment_arm*recon 0.5 0.5 -0.5 -0.5;
run;
But the log returned me with
"NOTE: trt diff is not estimable.". (no error message)
When I use the proc mixed, it works fine. What is the problem?
Thank you! | p******p
发帖数: 13 | 2 问了下自己这边的统计师，明确的random effect最好用proc mixed，proc glm貌似是
当成fixed effect处理的。 |
|
