> That's because variance of foo is measured in foo^2 units. What is
> the square of an interval?
>
> Cheers,
> D
Well if you're willing to accept that for the purposes of computing the
aggregates, an interval "month" is equal to 30 days (which is how
avg(interval) already works), then an interval is reducable to a single
quantity -- a number of seconds -- which can be squared.
30 days per month is pretty rough ... we could refine it to 30.4375, which
is the average number of days per month over four years including one leap
year.