From: William Shakespeare on
I have data that looks like this:

id cell pur_type buyer
1 1 shoes 1
1 1 other 1
2 1 shoes 1
3 2 shoes 1
4 2 shoes 1

I can write a proc tabluate that uses formats to generate a table like
this:


Group #Buyers
------------------------------------------
Cell 1 Shoes 2
Cell 1 Other 1
Cell 2 Shoes 2
------------------------------------------
Total 5

The problem is with the total of 5 buyers. Since id=1 has purchases in
two categories it is counted twice. This is fine if I want to compute row
percentages, but my bottom line counts are distorted.

I'd like something like this:

Group #Unique Buyers
------------------------------------------
Cell 1 Shoes 2
Cell 1 Other 1
Cell 2 Shoes 2
------------------------------------------
Total 4

This way, row percentages would have the correct denominator while the
total number of unique buyers would be accurate. Any suggestions on the
best way to accomplish this?