I am stuck in a problem, and I can't think of a next step to find the solution. The question is the following:

Suppose an urn has $k$ balls, numbered from $1$ to $k$, $k \in \mathbb{N}$. A sample of $m$ balls is withdrawn, $m \in \mathbb{N}^ \ : \ 1\leq m < k$ and the numbers of such sample are written down. Then the $m$ balls are returned to the urn and another sample is withdrawn, with $m$ balls again and registering the numbers. Let $X_n$ be the number of distinct numbers observed at least once in the first $n$ extractions, $n \in \mathbb{N}$. Find $E(X_n)$ and $Var(X_n)$. (Tip: $X_1 = m$)

Since the number of extractions will be at least $m$ and at most $k$, given that $X_1 = m$ I tried to evaluate $P(X_n = j + m | X_1 =m)$ for $0\leq j \leq \min\{m, k-m \}$. It turns out that $X_2$ have a hypergeometric distribution but I do not seem to find an answer trying to expand it for $X_n$ and I am really stuck.

Any tips or advices?

Thank you so much!


I think it will be much easier to go about this if we consider the random variables $Y_i,\ i=1,\dots,k$ where $Y_i$ is $1$ if ball number $i$ is drawn at some time in the first $n$ draws and $0$ if it is not. Then $X_n=\displaystyle{\sum_{i=1}^kY_i}$ and it's easy to compute $E(X_n)$ by linearity of expectation. The variance is harder because the $Y_i$ aren't independent, but you can use the formula for the variance of the sum in terms of the covariance.

  • $\begingroup$ Saul, I will follow your tip and edit the problem with a solution, to see if I get it right. Thanks! $\endgroup$ – M.Gonzalez Apr 19 at 21:09

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.