06-19-2013 05:06 PM
I am trying to create different samples of 0 (non event) from a large population in order to have a variety( a couple of thousands) of balanced samples ( equal obs of 0 and 1) to use as training and validation set, so in the end score the rest of population.
I would like the seed of generating the sample for 0s every time to change so as to eliminate the risk of having overlapping zeros across the sets.
My 1s should be the same in every set, i know that proc survey select should do the trick but i am stuck in the loop one could say.
Thank you in advance