Roger,

All too often the study of data requires care.

Given the sample size, even assuming the correctness of your
statistical model, the uncertainty in your estimate is so large
as to make your result worthless. IOW, estmating the tail of
the distribution at an 80,000 year return period from ~100 years
of data is simply silly. It's abuse like this that gives the field
of statistics a bad name. Please stop.

Cheers,
Russell

Roger Coppock wrote:
Atlantic TS Epsilon Is About a 1 in 80,000 Year Event!!!!!

Tropical storm Epsilon, number 29, is now active in the Atlantic:

http://www.wxforecasts.com/ameriwx/a...age&hwvmetric=

http://www.wxforecasts.com/ameriwx/a...ems&hwvmetric=

If one controls for the increasing capability to observe
tropical storms over the last 126 Atlantic hurricane seasons,
this year's 29 Atlantic storms are about a 1 in 80,000 event!

If one ignores the 4 storms per year per century growth in
counted storms, this year's 29 storms is a 1 in 1,700,000
event. I won't use this, because it is misleading. I'll
leave the lying and exaggeration to the Carbon industry
lobbyists. They are the experts at that.

Storms
[1] 11 7 6 4 4 8 12 19 9 9 4 11 9 12 7
[16] 6 7 6 11 9 7 12 5 10 5 5 11 5 10 11
[31] 5 6 7 6 1 5 14 3 5 3 4 6 4 7 8
[46] 2 11 7 6 3 2 9 11 21 11 6 16 9 8 5
[61] 8 6 10 10 11 11 6 9 9 13 13 10 7 14 11
[76] 12 8 8 10 11 7 11 5 9 12 6 11 8 8 15
[91] 10 13 7 8 11 9 10 6 12 9 11 12 6 4 13
[106] 11 6 7 12 11 14 8 7 8 7 19 13 8 14 12
[121] 15 15 12 16 15 29

The data are from the NOAA web page given at this URL:
http://www.aoml.noaa.gov/hrd/hurdat/...arandStorm.htm
Currently, these data are preliminary and dated Friday,
July 15, 2005 9:00PM.

sort(Storms)
[1] 1 2 2 3 3 3 4 4 4 4 4 4 5 5 5
[16] 5 5 5 5 5 5 6 6 6 6 6 6 6 6 6
[31] 6 6 6 6 6 7 7 7 7 7 7 7 7 7 7
[46] 7 7 7 8 8 8 8 8 8 8 8 8 8 8 8
[61] 9 9 9 9 9 9 9 9 9 9 9 10 10 10 10
[76] 10 10 10 10 11 11 11 11 11 11 11 11 11 11 11
[91] 11 11 11 11 11 11 11 12 12 12 12 12 12 12 12
[106] 12 12 13 13 13 13 13 14 14 14 14 15 15 15 15
[121] 16 16 19 19 21 29

Storms[126]
[1] 29
mean(Storms)
[1] 9.142857
sd(Storms)
[1] 4.086493

fitted.model

Call:
lm(formula = Storms ~ Year, data = aframe)

Coefficients:
(Intercept) Year
-70.80045 0.04115


residuals(fitted.model)[126]
126
17.28496

residuals(fitted.model)[126]/sd(Storms)
126
4.229779

pnorm(residuals(fitted.model)[126]/sd(Storms))
126
0.9999883

# Yes, except for this year's extreme and a few others,
# the residuals of the "Storms" data are normally
# distributed. A normal quartile-quartile plot shows a
# straight line, confirming this.

1/(1-pnorm(residuals(fitted.model)[126]/sd(Storms)))
126
85498.98
# Which rounds to 1 in 80,000.

# NOTE:
# to remove any bias due to increasing capability to observe
# hurricanes, I have used the current value of the trendline
# and not the mean as a baseline. Thus, the call to the
# residual here. Had I used the mean as a base line the
# result would have been 1 in 1,700,000. I'll leave the
# lying and exaggeration to the fossil fools.

pnorm((Storms[126]-mean(Storms))/sd(Storms))
[1] 0.9999994
1/(1-pnorm((Storms[126]-mean(Storms))/sd(Storms)))
[1] 1697027