6
$\begingroup$

I'm trying to find out the name of a distribution that is like negative binomial, only for finite population and without replacement. Or like Hypergeometric distribution where the last event has to be a success.

That is:

Let's say we have N balls in an urn, where W of them are white balls and B are black balls. I want to know what are the chances of needing to draw exactly n balls so it is the first time I got k of the n drawn balls to be black (e.g: that the last ball drawn was black, and a total of k out of n are black).

I can't seem to find the name of such a distribution, so if you could also advise me on where to find names of distributions, that would also be lovely.

Thanks.

$\endgroup$
3
$\begingroup$

You're thinking of the negative hypergeometric distribution. The top result in a search led to this description:

A negative hypergeometric distribution often arises in a scheme of sampling without replacement. If in the total population of size $N$, there are $M$ "marked" and $N-M$ "unmarked" elements, and if the sampling (without replacement) is performed until the number of "marked" elements reaches a fixed number $m$, then the random variable $X$ — the number of "unmarked" elements in the sample — has a negative hypergeometric distribution. The random variable $X+m$ — the size of the sample — also has a negative hypergeometric distribution.

$\endgroup$
3
$\begingroup$

I'm far not a distributional connoisseur, but it seems to me there is no need for a special distribution. Hypergeometric distribution is for sampling without replacement and will work here.

In your notation, $N$ is the population of balls containing $B$ black balls ("successes"). You draw sample of $n$ balls. The probability that $k$ of them are black (= the probability that sample sized $n$ is needed to contain $k$ black) is $Prob = PDF.HYPERGEOM(k,N,n,B)$.

$Prob$ is the probability that the sample is going to be drawn in any order of balls; there are $n!$ possible versions of order (permutations). Any unique ball could appear the last one (or the first one, or the second one - whatever you like) in $\frac{1}{n} n!$ versions. But you have $k$ black balls which are all the same for you in that respect. So, $\frac{k}{n}$ fraction of the order versions will have a black ball as the last drawn; and thus $\frac{k}{n} Prob$ should be the probability you ask.

$\endgroup$
  • $\begingroup$ Thanks. That is indeed true. And it seems that it is called Negative Hypergeometric Distribution, as was answered here: math.stackexchange.com/questions/355249/… $\endgroup$ – Tal Galili Apr 9 '13 at 6:35
  • 1
    $\begingroup$ Pah! @Tal, Did you know that cross-posting a question is strongly disapproved? ;) $\endgroup$ – ttnphns Apr 9 '13 at 7:45
  • $\begingroup$ I did :) I came across the math.SE after posting here, and decided they might have someone who would know (which they did...) But I'll avoid it is much as possible in the future :) $\endgroup$ – Tal Galili Apr 9 '13 at 11:48

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.