> My very rough sampling method came up with the same number
> you'd get by the usual method of doubling the official estimate
> (100,000), halving the organizers' (500,000) and striking an
> average. [225,000]
Maybe this is a peculiar instance, but using a geometric mean** yields a very similar number:
(100,000*500,000)^1/2 = 223,607
I appreciate that the usual method tries to account for the fact that cops tend to underestimate, while organizers tend to overestimate. But were there more specific reasons for working it as described above? I mean, does it have any advantage in these situations over using a geometric mean?
-- Shane
** - http://mathworld.wolfram.com/GeometricMean.html -------------- next part -------------- An HTML attachment was scrubbed... URL: <../attachments/20040830/5010c999/attachment.htm>