Archives of the TeradataForum

Message Posted: Wed, 30 Jan 2002 @ 13:25:22 GMT

	<Prev	Next>		<<First	<Prev	Next>	Last>>

I don't think sample or random can insure that you get rows from all groups. If you don't have to insure randomness, you could add a RANK to each row and GROUP BY your business groups. Then use QUALIFY to select rows with a RANK value less than the sample size you want divided by the number of groups.

Select USER_ID, GROUP_ID, RANK(USER_ID)
From TABLE_NAME
Group By GROUP_ID
Qualify RANK(USER_ID) <= (Sample Size / 20)

	<Prev	Next>		<<First	<Prev	Next>	Last>>