Beginning with SAS/STAT 9.22 in SAS 9.2 TS2M3, you can use the LSMEANS statement in PROC SURVEYREG to compute and compare least squares means (LS-means) of fixed effects.

Prior to SAS 9.2 TS2M3, you can compute least-squares means by using the ESTIMATE statement. Use the desired LSMEANS statement in PROC GLM, and specify the E option to generate the appropriate coefficients for use in the ESTIMATE statement in PROC SURVEYREG.

The following statements create data sets to be analyzed by PROC SURVEYREG. LSMEANS for the two states, Iowa and Nebraska, are desired.

data Farms;
   input State $ Region FarmArea CornYield Weight;
   datalines;
   Iowa     1 100  54 33.333
   Iowa     1  83  25 33.333
   Iowa     1  25  10 33.333
   Iowa     2 120  83 10.000
   Iowa     2  50  35 10.000
   Iowa     2 110  65 10.000
   Iowa     2  60  35 10.000
   Iowa     2  45  20 10.000
   Iowa     3  23   5  5.000
   Iowa     3  10   8  5.000
   Iowa     3 350 125  5.000
   Nebraska 1 130  20  5.000
   Nebraska 1 245  25  5.000
   Nebraska 1 150  33  5.000
   Nebraska 1 263  50  5.000
   Nebraska 1 320  47  5.000
   Nebraska 1 204  25  5.000
   Nebraska 2  80  11 10.000
   Nebraska 2  48   8 10.000
   Nebraska 3 180  13 10.000
   Nebraska 3 148  28 10.000
   ;
data TotalInStrata;
   input State $ Region _TOTAL_;
   datalines;
   Iowa     1 100
   Iowa     2  50
   Iowa     3  15
   Nebraska 1  30
   Nebraska 2  20
   Nebraska 3  20
;

Prior to the analysis by PROC SURVEYREG, fit the model in PROC GLM and include the desired LSMEANS statement with the E option to display the LSMEANS coefficients. The ODS SELECT statement selects only the table of LSMEANS coefficients for display

ods select LsMeanCoef;
proc glm data=Farms;
   class  State Region;
   model  CornYield = state region FarmArea State*Region;
   weight Weight;
   lsmeans state / e;
   run;
   quit; 

Now fit the model in PROC SURVEYREG and include an ESTIMATE statement for each LSMEAN that specifies the coefficients as provided by PROC GLM.

proc surveyreg data=Farms total=TotalInStrata;
   strata State Region / list;
   class  State Region;
   model  CornYield = state region FarmArea State*Region /solution;
   weight Weight;
   estimate 'Lsmean for State=Iowa'
            intercept 1 state 1 0 region .333333 .333333 .333333
            farmarea 130.666667 state*region .333333 .333333 .333333 0 0 0;
   estimate 'Lsmean for State=Nebraska'
            intercept 1 state 0 1 region .333333 .333333 .333333
            farmarea 130.666667 state*region 0 0 0 .333333 .333333 .333333;
   run;

   proc surveyreg data=Farms total=TotalInStrata;
      strata State Region / list;
      class  State Region;
      model  CornYield = state region FarmArea State*Region /solution;
      weight Weight;
      estimate 'Lsmean for State=Iowa'
               intercept 1 state 1 0 region .333333 .333333 .333333
               farmarea 130.666667 state*region .333333 .333333 .333333 0 0 0;
      estimate 'Lsmean for State=Nebraska'
              intercept 1 state 0 1 region .333333 .333333 .333333
              farmarea 130.666667 state*region 0 0 0 .333333 .333333 .333333;
      run;

Following are results from the ESTIMATE statements. Currently, confidence intervals for the estimates are not available.