2
$\begingroup$

Let's say, we have $i$ continuous IID random variables $X_1, X_2, \ldots,X_i$ whose domain is $\mathbb R$. This sequence divides the real axis into $i + 1$ intervals. Now, if we have another random $X_j$ who also have the same distribution, is correct that the probability of this $X_j$ falling into these $i+1$ intervals are equal?

(For me, it seems to be intuitively correct since there are no difference between these $X$'s. )

$\endgroup$
1
$\begingroup$

Yes. Think of the $(i+1)!$ orderings of the $i+1$ variables $\{X_j\}_{j=1}^{i+1}$. Each occurs with an equal probability of $\dfrac{1}{(i+1)!}$. $X_{i+1}$ is in position $j$ for exactly $i!$ of those arrangements. Thus, it is in each position with probability $\dfrac{i!}{(i+1)!}=\dfrac{1}{i+1}$.

$\endgroup$
  • $\begingroup$ I should mention that I did use the continuity of the random variables by assuming that the probability that any two variables were equal is $0$. $\endgroup$ – robjohn Sep 20 '11 at 22:20
1
$\begingroup$

On the basis that you have no more information about the distribution or indeed the values actually taken, then yes.

To see this, instead consider the $i+1$ random variables $X_1, X_2, \ldots X_i$ and $X_j$. Since they are identically and independently distributed, the $(i+1)!$ possible orders are equally likely, and the values will be distinct with probability $1$.

So it is equally likely that $X_j$ takes any of the $i+1$ possible positions in the order, and so it is equally likely that $X_j$ is in any of the $i+1$ intervals created by the other $i$ random variables.

$\endgroup$

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.