Josh Berkus <josh@agliodbs.com> writes:
>> I'm not convinced you can get a sufficiently good estimate from a small
>> subset of pages.
> Note that if this requires VACUUM rather than ANALYZE, it introduces a
> problem for data warehousing users, who can go years between vacuums of
> their largest tables.
It's likely that the default estimate of zero index correlation will
work just fine for such users ...
regards, tom lane