Suppose we've given an urn which contains $R$ red and $W$ white balls. These balls are drawn randomly from the urn and are not placed back. Let

  • $X:=$ number of attempts, before we've drawn at least $r\le R$ red balls
  • $Y:=$ number of red balls after $n-1$ attempts

I want to calculate $\text{Pr}(X=n)$.

Suppose we've already drawn $n-1$ balls from the urn and received $r-1$ red balls. The probability of this to happen is given by $$\text{Pr}(Y=r-1)=h(r-1|R+W,R,n-1):=\frac{\begin{pmatrix} R \\ r-1 \end{pmatrix}\begin{pmatrix} W \\ n-r \end{pmatrix}}{\begin{pmatrix} R+W \\ n-1 \end{pmatrix}}$$ where $h$ denotes the hypergeometric distribution. The probability that now another red ball is drawn is given by $$\text{Pr}(X=n)=\text{Pr}(Y=r-1)\;h(1|R+W-(n-1),R-(r-1),1)=\frac{\begin{pmatrix} R \\ r-1 \end{pmatrix}\begin{pmatrix} W \\ n-r \end{pmatrix}}{\begin{pmatrix} R+W \\ n-1 \end{pmatrix}}\frac{R-(r-1)}{R+W-(n-1)}$$ While I think that's correct, I would really like to get a more compact form of that. After some research on the internet, I found out, that I should be able to receive $$\text{Pr}(X=n)=\ldots =\frac{\begin{pmatrix} r-1 \\ n-1 \end{pmatrix}\begin{pmatrix} R+W-(r-1) \\ R-(n-1) \end{pmatrix}}{\begin{pmatrix} R+W \\ R \end{pmatrix}}$$ But I don't see how I get the $\ldots$ filled.

  • $\begingroup$ I'm unsure whether the final expression is correct or not. However, I'm sure we can rewrite the result in a more compact way. $\endgroup$ – 0xbadf00d Jun 8 '14 at 20:47

Your final expression may not be quite correct: for example I am dubious about ${r-1 \choose n-1}$ when $n \gt r$, but you can approach it from what you have done with something like this, which you should check and simplify. You also seem to have $K$ and $r$ representing the same thing.

$$\dfrac{\displaystyle{R \choose r-1}{W \choose n-r} }{\displaystyle{R+W \choose n-1} } \; \dfrac{R-(r-1)}{R+W-(n-1)}$$

$$=\dfrac{R!\; W!\; (n-1)! \;(R+W-(n-1))!\; (R-(r-1))}{(R+W)!\; (r-1)!\; (R-(r-1))! \; (n-r)!\; (W-(n-r))! \;(R+W-(n-1))}$$

$$=\dfrac{R!\; W!}{ (R+W)!}\; \dfrac{ (n-1)!}{(r-1)!\; (n-r)!} \;\dfrac{(R+W-(n-1))!}{(R-(r-1))! \; (W-(n-r))!}\dfrac{(R-(r-1))}{\; \;(R+W-(n-1))}$$

$$=\dfrac{\displaystyle {n-1 \choose r-1}\; {R+W-n \choose R-r}}{\displaystyle {R+W \choose R}}$$

though I have not checked this.

An alternative approach would be to find the probability of $r$ reds in the first $n$, i.e. $\dfrac{\displaystyle{n \choose r} {R+W-n \choose R-r}}{ \displaystyle{R+W \choose R}}$ and multiply this is the conditional probability that the last one of the $n$ is red, which I believe is $\dfrac{r}{n}$. I think this gives the same answer more quickly.

  • $\begingroup$ You're right. $K$ should be $r$, sorry for that. $\endgroup$ – 0xbadf00d Jun 8 '14 at 20:49

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.