> Levee height - 45 feet.
> Expected "average" water height - 36 feet.
> **Known** variation, during storms - +-15 feet

That's just what I was talking about. There is no such thing as a 
"known" variation. That 15 feet you mention (I know it's just an 
example) will represent some kind of statistical distribution. Does the 
water get to 36+15 feet *every* storm, 1 in 5 storms, 1 in 20 storms or 
what? The designers *will* have figured this out, in most developed 
countries it's illegal not to.

> This is a bit like playing Russian roulette, your "odds" might be 5 in 6
> of not getting the bullet, but.. if you do, then it doesn't bloody
> matter, at all, what the odds where, at that point of "not" getting it.

It does matter if you are expecting 500 storms over the next 50 years, 
it matters a lot whether you get flooded 250 times, 25 times, or twice.

> Yeah, there might have been, say, a 10% chance if the storm surging
> enough to make it over 45 feet, and a, say, 90% change that it would be
> "much" lower, but, it doesn't mean jack all if you are wrong, and jumps
> over into the 10% range of things that, "Probably won't happen."

If the engineer states that 90% of storms will not breach the levee, 
then he is not "wrong" when one does. If a statistically significant 
number of storms do breach the levee (ie not just 1 or 2 storms), and 
it's more than 10% *then* he is wrong.

> The actual, supposed, scientist, himself, simply ignored the odds of
> being wrong, and reported, "I see no reason to believe, at all, that it
> will rise higher than the 'average'."

And nobody questioned this ridiculous statement? Did this person 
actually have any qualifications?

> A book I read a while back stated that this was a bit like a common joke
> among statisticians: A statistician drowned in a river. The river had an
> 'average' yearly depth of only 0.5 inches." The point of the joke being
> that the river might be dead dry 3 weeks out of the year, and 50 feet
> deep, in a flash flood, during the middle of the rainy season, but the
> "average" is only 0.5 inches.

A statistician would obviously know to also look at the distribution and 
not just the average :-)

Post a reply to this message