In article , Torsten Brinch wrote:
On Sat, 22 May 2004 14:58:47 GMT, "Steve Young" wrote:
[...]
"They" say things like, That was a 100 year flood, expected to occur only
every 100 years. (or snow, drought, whatever) But I have doubts about how
they calculate that.

We do not need to get into the detail, but from a solid observation
series (preferably many hundreds of years :-) one can of course make a
good statistical estimate of what magnitude the event must have to
have only a 1 % chance of occurring in any given year. The problem
with some estimated 100 year events is that they do not have that kind
of solid base in observation. Say, if you have only 20 years worth of
data, you are thrust into making an extrapolation resting on a set of
assumptions, which reality may very well see fit to overturn in time.
[...]

Yeah. I recall sitting by a motel pool drinking beer late one evening
several decades ago and discussing "rainfall cycles" (you know the
sort of thing, the 11/13/whatever year "solar cycle" etc.). The local
mathematical statistician pointed out that you would need a minimum of
300 years of annual data to "see" such cycles with any confidence. As
he said, humans are always looking for patterns in things and are very
good at finding them, even if they are not real ones.

(Incidentally, I don't know if that "300 years" was just a figure
plucked out of the XXXX ambience, or whether he had some knowledge of
the distributions when he made the claim. As he was supporting a
large team of scientists working on pastoral systems at the time, it's
quite possible it was the latter.)


Cheers, Phred.

--
LID