Take the 2-minute tour ×
Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. It's 100% free, no registration required.

How do you show, that the expected value of a hyper-geometric random variable X with parameters $r$,$w$, and $n$ (a box contains $r$ red balls, $w$ white balls and $n$ balls are drawn without replacement) is $$\mathbf{E}(X)=\frac{rn}{r+w}?$$

I was able to reduce this to $$ \sum _{k=1}^{\infty}{r\choose k}{w\choose n-k}/{r+w\choose n} =\frac{n!(r+w-n)!r!w!}{(r+w)!k!} \sum _{k=1}^{\infty} \frac{1}{k!(r-k)!(n-k)!(w-n+k)!}. $$

No text I have seen yet proves this (and I think I know why mine doesn't either!). Any suggestions or references??

share|improve this question

2 Answers 2

up vote 3 down vote accepted

The linearity of expectation is the simple way to approach this problem. It is a very powerful technique that enables us to find the expectation of many random variables $X$ even when it is extremely difficult to find the distribution of $X$.

But if you really want to avoid using the linearity of expectation, it can be done in this case. The calculation will take a while: its length can be considered a proof of the fact we should use the linearity of expectation! By the general formula for expectation in the discrete case, when the distribution of $X$ is known, the expectation of the number of red balls $X$ is $$\sum_{k=0}^r kP(X=k).$$
The probability that $X=k$ is, as you know, $$\frac{\binom{r}{k}\binom{w}{n-k}}{\binom{r+w}{n}},$$ and therefore $$E(X)=\sum_{k=1}^r k \frac{\binom{r}{k}\binom{w}{n-k}}{\binom{r+w}{n}}. \qquad (\ast)$$ We are summing from $k=1$ on because the $k=0$ term makes no contribution to the expectation, and could cause some headaches later.

We use the following result:

Lemma: The binomial coefficient $\binom{q}{p}$ is equal to $\frac{q}{p}\binom{q-1}{p-1}$.

The lemma is easy to prove, either combinatorially or by manipulation. For a manipulational proof, note that $\tfrac{q}{p}\tbinom{q-1}{p-1}=\tfrac{q}{p}\tfrac{(q-1)!}{(p-1)!(q-p)!}=\tfrac{q!}{p!(q-p)!}$.

Using the lemma, we can see that $$\binom{r+w}{n}=\frac{r+w}{n}\binom{r+w-1}{n-1}.\qquad(\ast\ast)$$ We also need some information about $k\binom{r}{k}$. By the lemma, or otherwise, $$k\binom{r}{k}=r\binom{r-1}{k-1}.\qquad(\ast\ast\ast)$$ Substituting the values obtained in $(\ast\ast)$ and $(\ast\ast\ast)$ for the terms in the formula $(\ast)$ for the expectation of $X$, we obtain $$E(X)=\frac{rn}{r+w}\sum_{k=1}^r \frac{\binom{r-1}{k-1}\binom{w}{n-k}}{\binom{r+w-1}{n-1}}.$$ Make the change of variable $j=k-1$. Then the above formula for $E(X)$ becomes $$E(X)=\frac{rn}{r+w}\sum_{j=0}^{r-1} \frac{\binom{r-1}{j}\binom{w}{n-j-1}}{\binom{r+w-1}{n-1}}.$$

Note that $\frac{\binom{r-1}{j}\binom{w}{n-j-1}}{\binom{r+w-1}{n-1}}$ is the probability that when you draw $n-1$ balls from an urn that contains $r-1$ red and $w$ white, you will get exactly $j$ red balls. When we sum this from $j=0$ to $r-1$, we are adding up all the probabilities, so the complicated-looking sum is equal to $1$. We conclude that $$E(X)=\frac{rn}{r+w}\sum_{j=0}^{r-1} \frac{\binom{r-1}{j}\binom{w}{n-j-1}}{\binom{r+w-1}{n-1}}=\frac{rn}{r+w}.$$

Remark: Although the linearity approach is the smoothest, there are other properties of expectation that one can use for a proof. For example, let $E(n,x,y)$ be the expected number of red balls when we draw from $x$ red and $y$ white. On the first pick, we get a red with probability $\frac{r}{r+w}$, and a whie with probability $\frac{w}{w+r}$. If we get a red on the first pic, then our expected number of reds is $1$ plus the expected number of reds from the remaining picks. If we get a white, then our expected number of reds is simply the expected number of reds from the remaining picks. So we obtain $$E(n,r,w)=\frac{r}{r+w}(1+E(n-1,r-1,w))+\frac{w}{r+w}E(n-1,r,w-1).$$ Using this formula, and a simple induction on $n$, we can prove that $E(n,x,y)=\frac{nx}{x+y}$.

share|improve this answer

Use linearity of expectation. For each red ball, the probability of being drawn, and thus the expected number of successes produced by that ball, is $n/(r+w)$. By linearity of expectation, the expected number of successes is just the sum of the contributions from all $r$ red balls, i.e. $r\cdot n/(r+w)$.

share|improve this answer

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

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