proc mix: Reverse the direction of difference using PDIFF

[Arthur2]

Hi,

I'm using proc mixed as follows:

proc mixed data = mydata (where=(treatn in (1,2,3))) plots=all;
by estremn estrem atypen atype paramcd param avisitn avisit ;
class treatn;
model aval = treatn base/residual outp=resid;
lsmeans treatn / pdiff=control('3');
run;

The difference of lsmeans and SE for 2 treatments (treatn=1,2) is required to be estimated from placebo (treatn=3) as 'placebo - treatment'. I believe currently pdiff is giving the difference as 'treatment - placebo'.
Help is required to reverse the direction of the difference so that I can get the difference of each treatment from placebo.

Thanks

[ChrisNZ]

Changed title and moved to statistics community.

[FreelanceReinh · Accepted solution]

Hi @Arthur2,

You can use the LSMESTIMATE statement (instead of the PDIFF option of the LSMEANS statement) to request the statistics for these differences (or for more general linear combinations).

lsmeans treatn;
lsmestimate treatn 'placebo vs. trt 1' -1  0 1,
                   'placebo vs. trt 2'  0 -1 1;

[SteveDenham]

@FreelanceReinh , nice to see that your answer matches mine.  This thread is duplicated, and I am not sure how to merge the two.

SteveDenham

[FreelanceReinh]

@SteveDenham: That's funny. Rather, I am glad to see that my tentative idea matches a real expert's advice. 🙂

I wasn't aware of the duplicate thread. I think only moderators, super users and very few others have the privileges to merge threads.