> It certainly should be the case. starelid matches to pg_class.oid and
> staattnum matches to pg_attribute.attnum.
My problem was that I was looking up "event_date_time" in pg_class.relname
(and finding it), but the oid matched nothing. when I looked for 'event' in
pg_class & 'event_date_time' in pg_attribute, everything became clear.
So... I updated stadistinct to the actual value of 92,000 as you suggested
and as you predicted, it did not change the plan a bit. I guess I'll look
elsewhere for ways to optimize this query.
Thanks!
-Nick