0
$\begingroup$

I have three random variables $(X, Y, Z)$. The variables $X$ and $Y$ are conditionally independent given $Z$.

Will it be correct if I assume $P(X | Z) = P(X)$ or $P(X | Y) = P(X)$ ? If not, then why? Can you give a simple example to explain why the above assumption will not be correct?

$\endgroup$
1
$\begingroup$

Let $X:=\epsilon_1+Z$ and $Y:=\epsilon_2+Z$, where $\epsilon_1$ and $\epsilon_2$ are independent $N(0,1)$ (independent of $Z$) and $\mathsf{P}(Z=1)=\mathsf{P}(Z=-1)=1/2$. $$ \mathsf{P}(X\le 0\mid Z)=\mathsf{P}(\epsilon_1+Z\le 0\mid Z)=\Phi(-Z). $$ But $\mathsf{P}(X\le 0)=1/2$.


Conditional independence means that for all Borel sets $A$ and $B$, $$ \mathsf{P}(X\in A, Y\in B\mid Z)=\mathsf{P}(X\in A\mid Z)\mathsf{P}(Y\in B\mid Z). $$

$\endgroup$
  • $\begingroup$ Is $\Phi(Z)$ a cumulative probability function? And, is $Z$ symmetric about 0 ? $\endgroup$ – user10853036 Feb 5 '19 at 22:08
  • $\begingroup$ Yes, $\Phi$ is the cdf of $N(0,1)$. $\endgroup$ – d.k.o. Feb 5 '19 at 22:31

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.