You could make a string of the numbers and missing values end then use Proc Sort with Nodupkey. Best regards Per Andersen

Jennifer.Soller@sf.frb.org skrev i artiklen <869670453.14828@dejanews.com>... > Hello to all! I am once again searching for elegance. > > I have written a macro which generates a column of 50 numbers > and missing values placed in a specific order. The numbers were > initially randomly generated and then placed in their position > based on a set of convoluted constraints, with the missing value > serving as a delimiter between groups: > > OBS I1 > 1 . > 2 13 > 3 42 > 4 53 > 5 . > 6 19 > 7 . > 8 4 > 9 6 > 10 15 > 11 41 > > and so on, down to 62 places (the 50 states divided into 12 groups). > Right now, my macro generates this column (called I1 for iteration 1). > I am going to generate hundreds, if not thousands, of these columns > but I need to make sure that no column is generated more than once. > The end result would be a dataset containing different permutations of > 50 states assigned to 12 groups. So I need some way to compare the > columns as they are generated: > > upon generation of column I2, compare it to I1 -- if it matches, > delete > it. Generate I3, compare it to I1, delete if match, compare to I2, > delete if match, and so on. > > I've been brainstorming and have come up with a really convoluted, > messy way to do this -- most likely any way would be better. Anyone > up for the challenge? Or is this really simple, and I am just not > seeing it? > > Thanks in advance. > > Jennifer Soller > Senior Research Associate > Federal Reserve Bank of San Francisco > (415) 974-2377 Jennifer.Soller@sf.frb.org > > -------------------==== Posted via Deja News ====----------------------- > http://www.dejanews.com/ Search, Read, Post to Usenet >