Roger Coppock wrote:
"IOW, estmating the tail of the distribution at an
80,000 year return period from ~100 years
of data is simply silly" --- Russell

No, it is people who make statements without
either data or calculation to back them up that
gives the field of statistics a bad name.

You're wrong about that. People who make statements
without data to back them up are not doing statistics at
all. People who do statistics wrong or abuse the process
of drawing conclusions based on statistics are the people
who give it a bad name.

Do you have either numbers or calculation to back
your conjecture, Russell?

I don't need to do a calculation to know that, any more
than I would need to do a calculation to say that some-
one's claim of inventing a perpetual motion machine
is bogus. You should read the first few chapters of
_Statistics of Extremes_ by Gumbel. If you don't
believe me, try posting your result to sci.stat.math or
sci.stat.consult and see what they say. If you do I
suggest you pay particular attention to the comments
of Prof. Rubin and Reef Fish, if they reply, although
several other people there are quite knowledgeable.
I'd be happy to put my statement against yours and let
those more expert in statistics than either of us decide
which of us is closer to correct.

Just because you
are a bit giddy does not alter facts.

Extrapolating a model far beyond the range that the
data supports does not establish the facts, it just looks
like poor science. I'm not giddy, I'm the one who is
acting grounded. You're acting giddy if you think you
can say that an event observed once in 129 years of
imperfect data, or even perfect data, can be used to
*reliably* infer an 80,000 year return period. Instead
of arguing with someone like me, a scientist who wants
good science done in the process of establishing the
facts, you should spend your time learning about things
when they are pointed out to you. Read Gumbel, which
is a classic and IMO quite readable as statistics texts
go. If you really do understand the reasoning behind
statistics then you know that your stated conclusion is
far too strong. The event was rare and it *might* even
be about an 1 in 80,000 year event, but the data simply
are insuffcient to draw a conclusion that precise.

Cheers,
Russell