0
$\begingroup$

The earthquakes occur in a region in accordance with a Poisson Process, at a rate of seven per year. Assume that all required conditions for the Poisson process hold.

What is the probability that in exactly three of the next eight years no earthquakes will occur?

My answer: We look for parameter $\lambda$ for 3 years, so $\lambda_3 = 7*3 = 21$

Let X be no. of earthquakes that occur in 3 years, P(X=0) = e^(-$\lambda_3)$ = e^(-21)

Let Y be no. of years without an earthquake in 8 years, then Y is a binomial random variable, Y~Binom(N=8, p=e^(-21)).

Then, P(Y=3) = (8C3)((e^(-21))^3)(1-e^(-21))^5 = 2.4414*10^(-26)

May I know if this is correct?

$\endgroup$
2
  • $\begingroup$ No, it isn't correct. The right way to approach this is to first find the probability $p_0$ of a single given year containing no earthquakes; this uses the Poisson distribution. Then, find the probability that out of eight years, exactly three have no earthquakes; this uses the binomial distribution. I think your problem is that you've used the $3$ in two different places, where it should only be used in one. $\endgroup$
    – Brian Tung
    Commented Nov 18, 2022 at 5:37
  • $\begingroup$ @BrianTung Thank you! May I know if this logic applies to finding the probability that at most 3 of the next 30 years no earthquakes will occur by an appropriate Poisson distribution? $\endgroup$
    – preti1323
    Commented Nov 18, 2022 at 5:50

1 Answer 1

0
$\begingroup$

You have the right idea, but the wrong application.

To understand where you went wrong, note that if $Y$ is the random number of calendar years (i.e. non-overlapping time intervals from January 1 to December 31) in which no earthquakes occur, then for $8$ such years, $Y$ has a binomial distribution with $n = 8$ and some probability $p$ that represents the probability that no earthquakes occur in one calendar year.

Think of it this way: suppose at the beginning of each year on January 1, you decide to count how many earthquakes happen during that year, and at the very end of December 31 of that year, you tally up how many occurred. If no earthquakes occurred, you consider that year a "success." Each year's outcome is an independent Bernoulli trial, and the total number of years in which no earthquakes occurred is a binomial random variable.

So, while you are correct that $Y$ is binomially distributed, you incorrectly determined the success parameter $p$. The correct value should be the probability that, in a single year, no earthquakes occurred. Since their occurrences follow a Poisson process with rate $\lambda = 7$ per year, the random number of earthquakes in a year is $$X \sim \operatorname{Poisson}(\lambda = 7).$$ Hence $$p = \Pr[X = 0] = e^{-7}.$$

Then the probability that we observed exactly $Y = 3$ years without earthquakes in an $8$-year period is simply $$\Pr[Y = 3] = \binom{8}{3} p^3 (1-p)^{8-3} = 56 e^{-21} (1 - e^{-7})^5.$$ This is a very small number--approximately $1$ in $23.66$ million, but nowhere near as small as the one you obtained.

Your solution is calculating something different, which is the probability that there are exactly three three-year periods in which no earthquakes occur over a total observation time frame of $24$ years.

$\endgroup$
2
  • $\begingroup$ WRT your last paragraph: Don't those three-year periods also have to be three-year-aligned? $\endgroup$
    – Brian Tung
    Commented Nov 18, 2022 at 6:37
  • $\begingroup$ @BrianTung Yes, but the imprecision of the English language being what it is (at least, relative to the precision of mathematics), it is difficult to describe what I mean. Each binary outcome of "no earthquakes" or "at least one earthquake" is assessed over a contiguous three-year period that is non-overlapping with other three-year periods, for a total duration of $24 = 3\cdot 8$ years, and the probability of having three such (non-overlapping but not necessarily consecutive) contiguous three-year periods out of 8 is what the OP had calculated. $\endgroup$
    – heropup
    Commented Nov 18, 2022 at 7:16

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .