New opportunities for statistics collection in Teradata 14.0 - comment by carrie

Hi Nazy,
 
If you don't collect stats for each separate combination, then there will no histogram where the optimizer can see the number of uniques and the interval values.  So if you are using those different combinations of those columns for different joins it would be advisable to collect stats on them.
 
However, since there are overlaps in the columns represented, the optimizer should choose to scan the table one time for the collection of all three stats.  It will first collect stats on the broadest set (A,B,C) and use the spool from that collection for the (A,B) and the (A) stats.   But for that to happen you will happen to bundle all three statistics in the same collect statistics statement.  
 
Thanks, -Carrie