Let $X$ and $Y$ be random variables such that $X \vert Y=y$ is normal distributed as $N(y,1)$ and Y is a continues random variable with PDF:


for $0<y<1$

and $0$ otherwise.

Find $Var(X)$

My idea is to find the marginal PDF for X and from there it's easy to find the variance. Now since the conditional is normal distributed, I should be able to find the simulatanous PDF by multiplying the the conditional PDF by the marginal PDF for Y. However when I try to use integration to find the marginal PDF for X with limits 0 and infinity, I get an uncomputable integral.... Why is this happening and/or is there another easier method? Thanks in advance!

  • $\begingroup$ You can find the variance from $E(X)=E\, [E(X\mid Y)]$ and $E(X^2)=E\, [E(X^2\mid Y)]$. $\endgroup$ – StubbornAtom Jan 8 '19 at 14:58

Use law of total variance, $$\mathbb{Var}(X)=\mathbb{E}[\mathbb{Var}(X|Y)]+\mathbb{Var}(\mathbb{E}[X|Y])$$

| cite | improve this answer | |
  • $\begingroup$ Okay so E[Var(XIY)] must be E[1]=1 since we know it's normal distributed with variance = 1. But how do I calculate Var(E[XIY])...? Is it using LOTUS? And wouldn't I also need to use Y's PDF in some way? $\endgroup$ – CruZ Jan 8 '19 at 16:54
  • 1
    $\begingroup$ @CruZ You are almost there. Think what random variable $\mathbb{E}[X|Y]$ is, and find its variance. $\endgroup$ – kludg Jan 8 '19 at 17:04
  • $\begingroup$ Okay so E[XIY]=y so I just have to calculate Var(Y)=E[Y^2]-(E[Y])^2? And I can do that using LOTUS! Is that correct? $\endgroup$ – CruZ Jan 8 '19 at 17:21
  • $\begingroup$ Looks reasonable to me, I would solve it so. $\endgroup$ – kludg Jan 8 '19 at 17:24

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.