On 17/03/2011 08:20, Chris Hogg wrote:

If an earthquake of this magnitude had a one-in-a-thousand-years
probability (not sure if that's actually the case in this instance),
and the plant is 40 years old (1970's vintage IIRC), dose that bring
the probability of it being hit by a really bad earthquake down to 1
in 25?

Not quite, although it doesn't alter your conclusion. If the 'quakes
are randomly distributed in time the Poisson distribution will apply to
finding the probability of any given number of occurrences in a
specified time interval.

P(k,lambda) = (lambda^k) * exp(-lambda) / k!

where lambda is the average no. of occurrences in the interval and k is
the particular no. of occurrences.

Here lambda is 0.04, so exp(-lambda) = 0.961, so the probabilities work
out as:

- no occurrences, P(0,0.04) = 0.961
- one occurrence, P(1,0.04) = 0.038 (1 in 26)
- two occurrences, P(2,0.04) = 0.00077 (1 in 1301)


--
Andy