1
$\begingroup$

I have a random variable denoted $N$. Then I have random variables denoted $X_1, X_2, \dots, X_N$ distributed according to a uniform distribution. I have also random variables denoted $Y_1, Y_2, \dots, Y_N$ distributed according to this same probability distribution.

If we consider the variable $S_X$ such that $S_X=\sum_{i=1}^NX_i$ and $S_Y$ such that $S_Y=\sum_{i=1}^NY_i$.

Are $S_X$ and $S_Y$ mutually independent ?

If we extend this problem to the case of any number $A$ of sums, for instance $S_X$, $S_Y$ and $S_Z$ when $A=3$. Are these $A$ variables (these 3 sums) mutually independent ?

Thank you.

$\endgroup$
  • $\begingroup$ Ok, thank you very much. In fact you could write this partial answer. I think this intuition is sufficient. $\endgroup$ – Dingo13 Jul 10 '13 at 17:05
  • $\begingroup$ rewrote it as you suggested. $\endgroup$ – gt6989b Jul 10 '13 at 17:06
  • 1
    $\begingroup$ Suppose the $X_i,Y_i$ are uniform on $[0,1]$ and we know that $S_X$ has taken on value $2013$. That makes it quite unlikely that $S_Y \lt \frac{1}{10}$! $\endgroup$ – André Nicolas Jul 10 '13 at 17:09
0
$\begingroup$

Assume $X_i,Y_i$ are already pairwise independent and let $\mathbb{E}[N] = \nu$.

Assuming $N$ is independent of the $X_i,Y_i$, Wald's indentity forces $$ \mathbb{E}[S_X] = \mathbb{E}[X_1] \mathbb{E}[N] = \nu/2 = \mathbb{E}[Y_1] \mathbb{E}[N] = \mathbb{E}[S_Y]. $$ There are similar results for other moments, so there is an explicit dependence on the distribution of $N$ for both $S_X$ and $S_Y$.

You can see that knowing the value of $S_X$ makes certain values of $N$ more likely than others, which will have an influnence on the distribution of $S_Y$.

On the other hand, fix $N$ to be a constant (not a random variable) and the dependence disappears.

$\endgroup$
  • $\begingroup$ Thank you for the prompt answer. $\endgroup$ – Dingo13 Jul 10 '13 at 17:07

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.