GLIMMIX - negative estimate

[rmn0008]

Hello. 

I'm having trouble with the estimates of LS-means from PROC GLIMMIX. Why are they negative?

PROC GLIMMIX data = COTreniform_1 plots = residualpanel;
class CULTIVAR NEMATICIDE NEMATODE REP;
model KG_HA = NEMATICIDE|NEMATODE / ddfm = satterth;
random YEAR|CULTIVAR;
lsmeans NEMATICIDE|NEMATODE / cl lines adjust = Tukey alpha = 0.05;
run;

[PaigeMiller]

They are negative because the fitted model says that the estimated least squares mean KG_HA for at least one of the levels of NEMATICIDE is negative (or similarly for estimated mean KG_HA for at least one of the levels of NEMATODE)

Why shouldn't they be negative?

[rmn0008]

PaigeMiller, 

Thank you for your input but I think that I found out what was the problem. 

I was not counting YEAR in the class statement. Now it's working!