Hi
I ran a program in both GLIMMIX and MIXED procs to perform Games-Howell test, but the software did not adjust DFs and the results are same as when I do not assume heterogeneity of variances.
This is the statements.
proc glimmix;
class trt;
model Cla=trt/ddfm=rk;
lsmeans trt/adjust=tukey adjdfe=row;
run;quit;
I appreciate any help.
Regards
It may just be a typo, but I assume you are applying a Kenward-Rogers adjustment for the denominator degrees of freedom adjustment (where you have ddfm=rk, I think it is probably ddfm=kr in your actual code). Unless you have unbalanced data, or an R side effect, the degrees of freedom will not change, so far as I remember. Are you comparing (homogeneous variance):
proc glimmix;
class trt;
model Cla=trt/ddfm=kr;
lsmeans trt/adjust=tukey adjdfe=row;
run;
to (heterogeneous variances by trt):
proc glimmix;
class trt;
model Cla=trt/ddfm=kr;
random _residual_/group=trt;
lsmeans trt/adjust=tukey adjdfe=row;
run;
Steve Denham
It may just be a typo, but I assume you are applying a Kenward-Rogers adjustment for the denominator degrees of freedom adjustment (where you have ddfm=rk, I think it is probably ddfm=kr in your actual code). Unless you have unbalanced data, or an R side effect, the degrees of freedom will not change, so far as I remember. Are you comparing (homogeneous variance):
proc glimmix;
class trt;
model Cla=trt/ddfm=kr;
lsmeans trt/adjust=tukey adjdfe=row;
run;
to (heterogeneous variances by trt):
proc glimmix;
class trt;
model Cla=trt/ddfm=kr;
random _residual_/group=trt;
lsmeans trt/adjust=tukey adjdfe=row;
run;
Steve Denham
Thank you dear Steve for the answer.
yes that was Kenward-Rogers.
I missed the random statement. I am comparing heterogeneous variances with unbalanced data. Can you explain a little what is R side effect, please?
regards
R side effects give rise to marginal estimates, and are the equivalent of the REPEATED statement in PROC MIXED. They are invoked in PROC GLIMMIX by including the residual option after the slash, or when only fitting for overdispersion/hetereogeneity with the _residual_ option before the slash.
G side effects give rise to conditional estimates, and are the equivalent of the RANDOM statement in PROC MIXED.
Steve Denham
Are you ready for the spotlight? We're accepting content ideas for SAS Innovate 2025 to be held May 6-9 in Orlando, FL. The call is open until September 25. Read more here about why you should contribute and what is in it for you!
Learn the difference between classical and Bayesian statistical approaches and see a few PROC examples to perform Bayesian analysis in this video.
Find more tutorials on the SAS Users YouTube channel.