2
$\begingroup$

I have this problem:

Let $X = V + W$ and $Y = V + Z$ where $V, W, Z$ are independent Pois($\lambda$) random variables.

I found that $Cov(X, Y) = Var(V) = \lambda$

It now asks to find whether $X$ and $Y$ are conditionally independent given $V$.

I am now trying to find if $X$ and $Y$ are conditionally independent given $V$.

So I start with $P(X = x, Y = y | V = v) =$

Now, I know that, for conditional independence, I have to show that the joint probability mass function factors into the product of its marginal probability mass functions.

But where do I go from here? This is where I have been stuck.

Please show me how this is done. Thank you very much for your help!

$\endgroup$
  • $\begingroup$ Maybe use that $X=x$ is the same as $V+W = x$ which is the same as $W = x-V$, do the same for $Y$. $\endgroup$ – Jesper Hybel Oct 16 at 10:34
  • $\begingroup$ @JesperHybel $P(X = x, Y = y | V = v) = P(W = x - v, Z = y - v | V = v)$ Hmm, I think it is $P(X = x, Y = y | V = v) = P(W = x - v, Z = y - v | V = v) = P(W = x - v, Z = y - v)$ because we already are given $v$, so the $| V = v$ part is redundant, right? $\endgroup$ – Wyuw Oct 16 at 10:43
  • $\begingroup$ From here I would just state that W and Z are independent so the P_{Z,W}(.) factorize as shown by @gunes $\endgroup$ – Jesper Hybel Oct 16 at 10:55
1
$\begingroup$

Intuitively, if $V$ is known, the only random component in $X$ will be $W$ and, in $Y$, it will be $Z$. Since the two are independent, so are $X$ and $Y$. More mechanically, you can start from your and @Jesper's comments: $$\begin{align}P(W=w,Z=z|V=v)&=\frac{P(W=w,Z=z,V=v)}{P(V=v)}\\&=\frac{P(W=w)P(Z=z)P(V=v)}{P(V=v)}\\&=P(W=w)P(Z=z)\end{align}$$

$\endgroup$
  • $\begingroup$ Ahh. At least I was on the right track. What rule is used to get $P(W=w,Z=z|V=v)=\frac{P(W=w,Z=z,V=v)}{P(V=v)}$? Some form of Bayes' rule? I am unsure of what rule was used to get each of the equality. Sorry, I am just learning this. $\endgroup$ – Wyuw Oct 16 at 10:56
  • $\begingroup$ It is the definition of conditional probability, i.e. $$P(A|B)=\frac{P(A,B)}{P(B)}$$ in its simplest form. $\endgroup$ – gunes Oct 16 at 10:57
  • $\begingroup$ Oh ok. But what was used for the separation $P(W=w,Z=z,V=v)=P(W=w)P(Z=z)P(V=v)$? $\endgroup$ – Wyuw Oct 16 at 11:01
  • $\begingroup$ That comes from the independence assumption stated in the question. $\endgroup$ – gunes Oct 16 at 11:02
  • $\begingroup$ Ahh, this all makes sense to me now! Thank you very much!!! $\endgroup$ – Wyuw Oct 16 at 11:03

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.