Longitudinal analysis - Proc Glimmix - LSMeans - Slice

[Rodolfo_Mingoti]

Hello, using the PROC GLIMMIX - LSMEANS I would like to get the differences between the groups (A, B, C and D) for each time (1, 2, 3 and 4).

Exemple:

Only for Time = 1

Tukey-Kramer Grouping for Grupo Least Squares Means (Alpha=0.05)
LS-means with the same letter are not significantly different.
Grupo	Estimate
Grupo D	1.8	A
Grupo B	1.7	B
		B
Grupo A	1.7	B
		B
Grupo C	1.6	B

 Only for time 2:

Tukey-Kramer Grouping for Grupo Least Squares Means (Alpha=0.05)
LS-means with the same letter are not significantly different.
Grupo	Estimate
Grupo B	1.9	A
Grupo C	1.5	B
		B
Grupo D	1.3	B
		B
Grupo A	1.0	B

Only for time 3 ...

The same idea...

Only for time 4 ...

The same idea...

My line commands were showed below:

PROC GLIMMIX DATA=WORK.Example;
	CLASS ID Group Time;
	MODEL var_study = Group Time Group*Time / DDFM=KR DIST=GAUSSIAN;
	RANDOM Time / RESIDUAL TYPE=CS SUBJECT = ID;
	LSMEANS Time*Group / LINES ADJUST=TUKEY;
RUN;
QUIT;

/* Dosen´t work the sintaxe = Lsmeans Group*Time / Slice=Time adjust=tukey;

So, in order to get the differences between the groups (A, B, C and D) for each time (1, 2, 3 and 4), how do I construct the LSMEANS procedure?

Thanks a lot,

Mingoti, Rodolfo.

[StatsMan]

Take a look at the SLICE statement in the PROC GLIMMIX documentation.  The syntax 

slice group*time / sliceby=time lines;

should give you what you are looking for.  In earlier releases of SAS, you will get the letters next to the lsmeans, indicating significance in the comparisons of the adjusted means.  In the latest release, you will get an ODS graphic that shows the significance across the comparisons.  These comparisons will be done within each level of the SLICEBY effect.  Add ADJUST=TUKEY if you want to get adjusted p-values within each level of the SLICEBY effect.  

[Rodolfo_Mingoti · Accepted solution]

Thanks a lot StatsMan!

I used your suggestion and got the results you needed.
Problem solved!

The final sintaxe was:

PROC GLIMMIX DATA=WORK.Example;
	CLASS id Group Time;
	MODEL var_study = Group Time Group*Time / DDFM=KR DIST=GAUSSIAN;
	RANDOM Time / RESIDUAL TYPE=CS SUBJECT = id;
		
SLICE Group*Time / SLICEBY=Time LINES ADJDFE=ROW ADJUST=TUKEY PLOT=NONE ;

RUN;
QUIT;