1
$\begingroup$

Given the following two R.V.s

$$z_{1} = \frac{x_{1}}{|x_{1}|^2 + |x_{2}|^2 + ... + |x_{M}|^2}$$

and

$$z_{2} = \frac{x_{2}}{|x_{1}|^2 + |x_{2}|^2 + ... + |x_{M}|^2}$$

where $x_{i} \sim \mathcal{CN}(0,a), \forall i$ and $a > 1$. As can be seen, the denominator follows a Chi-square distribution with $2M$ degrees of freedom as $x_{i}$ are i.i.d. R.V.s.

Is it possible to calculate

$$\mathbb{E} \{ z_{1} z_{2}^{*}\},$$

and show that the variables are correlated or not? Not that $*$ is the conjugate.

$\endgroup$

1 Answer 1

2
$\begingroup$

First here, by the Cauchy--Schwarz inequality, $$E|z_1z^*_2|=E|z_1|\,|z^*_2|\le\sqrt{E|z_1|^2E|z^*_2|^2}=E|z_1|^2<\infty,$$ by Addition in response to the modification of the OP's original question. So, $Ez_1z^*_2$ exists and is finite. Therefore and because the joint distribution of the pair $(-z_1,z^*_2)$ is the same as that of $(z_1,z^*_2)$, we conclude that $$Ez_1z^*_2=E(-z_1)z^*_2=0.$$

Detail sdded in response to the OP's comment: The $x_i$'s are iid and, for each $i$, the distribution of $-x_i$ is the same as that of $x_i$. So, the joint distribution of $(-x_1,x_2,\dots,x_M)$ is the same as that of $(x_1,x_2,\dots,x_M)$. Also, $|-x_1|=|x_1|$. So, the random pair \begin{multline*} (z_1,z^*_2)=g(x_1,x_2,\dots,x_M):= \\ \Big(\frac{x_{1}}{|x_{1}|^2 + |x_{2}|^2 + \dots + |x_{M}|^2}, \frac{x_{2}}{|x_{1}|^2 + |x_{2}|^2 + \dots + |x_{M}|^2}\Big) \end{multline*} equals \begin{multline*} (-z_1,z^*_2)= \Big(\frac{-x_{1}}{|-x_{1}|^2 + |x_{2}|^2 + \dots + |x_{M}|^2}, \frac{x_{2}}{|x_{1}|^2 + |x_{2}|^2 + \dots + |x_{M}|^2}\Big) \\ =g(-x_1,x_2,\dots,x_M) \end{multline*} in distribution, as was stated above.

$\endgroup$
8
  • $\begingroup$ Dear, thanks. I don't understand the inequality, is this some known one? Could you please, give further details on that? $\endgroup$ Feb 2, 2019 at 9:44
  • 1
    $\begingroup$ The only inequality in the above answer is an instance of the Cauchy--Schwarz inequality, which is now made explicit in the answer. $\endgroup$ Feb 3, 2019 at 1:49
  • $\begingroup$ I've accepted your answer as the correct one as I have simulated the correlation and it is zero as you pointed out, however, I fail to see how you conclude that the joint distribution of the pair (βˆ’π‘§1,π‘§βˆ—2) is the same as that of (𝑧1,π‘§βˆ—2) once the joint pdf is unknown. Is it possible to conclude that without knowing the joint pdf of (𝑧1,π‘§βˆ—2)? Thanks! $\endgroup$ Feb 3, 2019 at 11:17
  • 1
    $\begingroup$ @FelipeAugustodeFigueiredo : I have added the detail you requested. $\endgroup$ Feb 3, 2019 at 15:34
  • $\begingroup$ Thanks, it is clear now. $\endgroup$ Feb 3, 2019 at 17:23

Your Answer

By clicking β€œPost Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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