proc genmod data= ;

class ;

model y = x / dist= link= type3 ;

repeated subject= / corr= type= ;

lsmeans x / diff=control ('ref') cl adjust=Dunnett(or tukey);


distributionlink functionnote
normalidentitycontinuous dependent variable, mean difference
binomiallogrisk ratio from exponentiation of the parameter estimate 
binomiallogitodds ratio
poissonlogrisk ratio from exponentiation of the parameter estimate; used for robust error estimate with repeated subject statement

Parameters estimated with maximum likelihood methods. 

Repeated statement: specifies the covariance structure of multivariate responses and iterative fitting algorithm for GEE model fitting.