You could manipulate your data so that the values of your boxes are in alphabetical order, and then create a format so that what displays is still the meaningful names, like 'Nt', 'Nc', and 'Nh'. For example: data box;   do x='V3', 'V1', 'V2';     do i=1 to 10;       y=10*ranuni(2);       output;     end;   end; run; proc format; value $myboxval           'V1' = 'Nt'           'V2' = 'Nc'           'V3' = 'Nh'; run; proc sort data=box; by x; run; ods listing; ods graphics / reset width=5in height=3in imagename='BoxOrder'; footnote ''; proc boxplot data=box;   format x $myboxval.;   plot y*x=i;   run; ... View more

In your query, you could calculate a new variable using the INTNX function, and use that for your grouping instead of the original variable. In your case, intnx('month',work_date,0) Where 'month' is the interval type, and 0 is the interval since you want the same calendar month as the variable you started with. Note, work_date must be a date, not a datetime variable. The result will actually be a value equal to the 1st of each month, but then you can also apply the format you chose (mmyyyy) to this calculated variable so the output doesn't show the day of month, just month and year. ... View more