For fixed effects you should be able to use (treat is your factor name)

means treat;

lsmeans treat /pdiff;

This gives the means and the differences between pairs of means

the other is to request an orthogonal comparison where you

have to use:

estimate ' compare A and B' treat 1 -1 0 ;

the names of the treatments, and the coefficients are set up for your particular question.

For the random effects you get the variance components

lsmeans treat / adjust=scheffe;

other possibilities: BON DUNNETT NELSON SIDAK TUKEY T

If you have a mixed design (between x within) see attachment.

To All the folks above,

Greatly appreciate your inputs and the attached syntax!