Ashley, Again, moving in the right direction.  Cynthia's advice was good general advice, but you won't need to define %GLOBAL variables since you already have the & _COUNT variables to work with.  (At least probably, you won't need to.) As Art mentioned, there are some extra characters here and there.  Besides the periods after the semicolons, there are also a few %PUT A pieces to be removed. Other factors:  &DISTRICT_COUNT is being used to control iterations for &NOS, instead of &NOS_COUNT. Finally, the WHERE statement will not let you use _N_.  I recognize this as trying to simplify whether to add the word AND in later pieces of the WHERE statement.  I'm not sure if this will work either, but try this in a DATA step: data new; set old; where 1; run; If it works, that would be the way to start the WHERE statement (omitting the semicolon until you reach the end of the WHERE statement, as your code currently does).  If not, you might have to select a variable that is always nonmissing and condition on that, such as: where (district > ' ') ... View more

My guess: %sysfunc in combination with catx is calling %sysevalf in a way that we cannot see.  It's a sort of a black box.  We can see what goes in and what comes out, but can only guess at the internal workings. ... View more

Art, I like it.  So if you have a variable called _NAME_, is it implied that PROC TRANSPOSE should use it as an ID variable? ... View more

JohnK, In theory, it's not so hard.  In practice, though, you mention that your data set is long.  I suspect that "Firm" is not really as simple as A, B, and C, but takes on many more complex values that don't easily translate into unique variable names.  For example, if your data contained both of these Firms: George Washington Carver High School North George Washington Carver High School South I doubt that you could come up with a good way to assign what the new variable names would be. If you were willing to come up with a modified solution, it might look look like this.  Instead of variable names like A1, A2, B1, B2, use names like Sales0001_1, Sales0001_2, Sales0002_1, Sales0002_2, etc.  The first 4 digits of the new variables indicate which Firm it is, and you would need a separate method to translate from the arbitrary firm number to the firm name.  (This would not be difficult using a format.) Does this sound like an acceptable solution? ... View more

Ruth, I imagine there's not much out there in the way of macros, because SAS handles the problem easily enough without macros. If you can work with the number (instead of percentage) of missing values, it's as easy as this: proc means data=mydata maxdec=0 nmiss; run; If you really must have percentages, then create an output data set: proc means data=mydata noprint;    var _numeric_;    output out=missings (drop=_type_) nmiss=; run; In your output data set MISSINGS, each variable name will contain the number of observations that the original variable was missing.  You will also have the automatic variable _FREQ_, which is the total number of observations.  If you bring this into a DATA step, it will be easy enough to calculate the percentages. Good luck. ... View more