$\newcommand{\P}{\mathbb{P}}$$\newcommand{\E}{\mathbb{E}}$$X \sim N(0,1)$

$W$ is Rademacher distribution

$Y = WX$


In order to prove dependece of random variables I need to show, that $\P(X,Y) = \P(X)\P(Y)$ is violated.

I tried to find counterexample that would show, that the variables are dependent. First of all, I thought to show that $\E(XY) = \E(X)\E(Y)$ is not true. But in this case it is true $(\E(XY) = \E(X)\E(Y) = 0)$.

How can I prove, that variables are not independent

  • $\begingroup$ Could you clarify that you mean $Y=WX$ and not $Y$ is the convolution of $W$ and $X$? Further to this, this seems like a self study question, so if it is could you add the self study tag? $\endgroup$ – epp Nov 13 '16 at 12:39
  • $\begingroup$ @StatsPlease edited $\endgroup$ – Daniel Yefimov Nov 13 '16 at 12:44
  • $\begingroup$ If you need more help just comment on my answer and I'll attempt to guide you. $\endgroup$ – epp Nov 13 '16 at 12:54

What values can $Y$ take if $X=0$? What values can $Y$ take if $X=1$?

Are the conditional pdfs for $Y$ the same at the two values of $X$?

  • $\begingroup$ If $X = 0$, than Y can take only $0$ value. If $X = 1$, than Y can take values $1$ or $-1$ with $P = 1/2$ for both values. And from the fact above, i think, we can conclude, that pdfs for $Y$ at different values of $X$ are different. $\endgroup$ – Daniel Yefimov Nov 13 '16 at 13:15
  • 1
    $\begingroup$ @DanielYefimov But can you follow up your conclusion that the conditional distributions are different to arrive at the conclusion that $X$ and $Y$ are dependent random variables? $\endgroup$ – Dilip Sarwate Nov 13 '16 at 18:22
  • $\begingroup$ @DilipSarwate So we concluded, that our conditional distributions are different. It means, when variable $X$ has different values, the output of $Y$ changes. So, change in $X$ provoke change in $Y$. So, $X$ and $Y$ are dependent. Am I right? $\endgroup$ – Daniel Yefimov Nov 13 '16 at 18:31
  • 1
    $\begingroup$ @DanielYefimov Perfect! $\endgroup$ – Dilip Sarwate Nov 13 '16 at 20:05

I'll try and guide you to an answer.

We know:

$$X\sim N(0,1)$$


$$W=\begin{cases} \,\,\,\,\,1, & p=0.5\\ -1, & p=0.5\\ \end{cases}$$

We also know that $Y=WX$. So:

$$Y=\begin{cases} \,\,\,\,\,X, & p=0.5\\ -X, & p=0.5\\ \end{cases}$$

This shows that $Y$ is dependent on $X$.

  • $\begingroup$ Yes, it is known. $\endgroup$ – Daniel Yefimov Nov 13 '16 at 12:54
  • $\begingroup$ This answer started out nicely but the conclusion (everything after the dotted line) is sheer nonsense. It is not true that $Y = X$; what is true is that they have the same distribution. -1 pending correction. $\endgroup$ – Dilip Sarwate Nov 13 '16 at 14:57
  • $\begingroup$ @DilipSarwate Apologies for my answer. I'm not sure where the mistake is (I'm guessing it's a technical one). As far as I can see, it's reasonable to say $Y=X\,\,\text{w.p. } 1$ and they are perfectly dependent. I'm obviously missing something, so your help would appreciated. $\endgroup$ – epp Nov 13 '16 at 23:43
  • 1
    $\begingroup$ @StatsPlease "$X$" is a function. It is a special kind of function that we call random variable, because its range is characterized by another function, the "distribution function" that $X$ "obeys". We never write $X=N(0,1)$ or for any other distribution, namely we never use the $=$ symbol, and this is because $X$ follows a (in this case) Normal distribution, it is not "equal" to a normal distribution. In most cases we do not have a clue about the functional form of $X$. $\endgroup$ – Alecos Papadopoulos Nov 14 '16 at 19:36

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.