topic Re: PROC MIXED: specifying control in longitudinal model in Statistical Procedures

PROC MIXED: specifying control in longitudinal model

Altal — Mon, 28 Jul 2014 18:55:15 GMT

I'm working on a longitudinal model like the one below, with trt having 4 arms (1,2,3 and 4)

What is the syntax to have group 3 to be always the control in all visit*trt interactions in the lsmeans to get the difference?

proc mixed data=visits noclprint empirical;

class ID trt visit;

model dependant= visit trt visit*trt;

repeated visit / subject=ID(trt);

lsmeans visit*trt/diff pdiff ;

ods output LSMeans=means;

ods output Diffs=difference;

run;

Please for your help....

Re: PROC MIXED: specifying control in longitudinal model

Altal — Tue, 29 Jul 2014 12:17:42 GMT

I wasn't successful in getting the output I want using CONTROL option; however, I changed coding of arms so that 3 becomes 4 and 4 becomes 3; in this case output was always

1-4

2-4

3-4

which is what I was interested in.

any help regarding syntax without changing coding of groups would be appreciated.

Re: PROC MIXED: specifying control in longitudinal model

SteveDenham — Tue, 29 Jul 2014 13:14:55 GMT

You might look into the LSMESTIMATE statement to specifically look into comparisons.

Steve Denham