10-28-2013 08:10 AM
I want to reduce a really big data-set to fewer observations, but containing all the levels of the original data-set (at most) once. This would be done for testing purposes, so it's in our best interest to come up with final data-set that has least number of obs.
So if we take data-set SASHELP.CLASS for example. I would like to find out the first obs (or any non-specific obs really, but i feel it might be easier with first/last obs) that covers each individual levels of the original data-set, not all possible (existing, or theoretical) combinations of the required variables.
Output would look something like this: (SASHELP.CLASS: all levels by AGE, SEX):
So we should end up with:
Barbara, Carol, and Henry will not be output, as both F-M and 13-14 are covered
end of output
In this case we went through data sequentially, and we've found minimum number of obs (6) satisfying the requirement, but if the levels get more lengthy (or inter-correlated), and data assorted (supposed we could pre-sort), we might end up with something that is close to minimum, but not quite minimum of longest level among required variables.
I would imagine this would require some sort of recursive algorithm to get most efficiently small number of obs to cover the values, but have no clue where to start. Any help will be greatly appreciated!
10-28-2013 08:50 AM
This seems to work. You might replace the arrays with associative array so you don't have to guess at the dimension.
10-28-2013 10:19 AM
Thanks for your reply, data_null_;
It may be because of v9.1.3 of SAS i'm running, but I seem to be getting the following error after "fs" and "fa" definitions:
ERROR: The ARRAYNAME