Shelby Cain <alyandon(at)yahoo(dot)com> writes:
> Increasing the statistics target on the city column
> and re-analyzing the table seems to make the
> correlation estimate better (shows about 0.5) but the
> row count estimates are thrown off by 2 orders of
> magnitude in some cases.
What did you increase it to, exactly? Could we see the contents of
pg_stats for these two columns at both target settings?
regards, tom lane