While I wholeheartedly love and follow Mark's elegant and fast solutions, I find the above mine suits for lazy people like me and also handles wiiideeee and long datasets conveniently. There isn't much logic here to add my convenience except the diligence of array grouping i.e the 2nd grouping
On the one hand Proc mean does the means for var lists that I don't need to type hence you see double dash and on the other hand the datastep merely computes min for the lists.
Vname merely gets the variable name. No big deal about that.
EDIT: Also, this should handle multiple vars