0
$\begingroup$

Given three random variables $X_1,X_2,Y$ such that $X_2,Y$ are independent, how does one prove that

$$ E[g_1(X_1)g_2(X_2)|Y] = E[g_1(X_1)|Y]E[g_2(X_2)] $$

Attempt:

\begin{equation} \begin{aligned} E[g_1(X_1)g_2(X_2)|Y=y] =& \int \int_{\mathbb{R}^2} g_1(x_1)g_2(x_2)f_{X_1,X_2|Y}(x_1,x_2|y) dx_1dx_2\\ =& \int \int_{\mathbb{R}^2} g_1(x_1)g_2(x_2)\frac{f_{X_1,X_2,Y}(x_1,x_2,y)}{f_Y(y)} dx_1dx_2\\ =& \int \int_{\mathbb{R}^2} g_1(x_1)g_2(x_2)\frac{f_{X_1|X_2,Y}(x_1|x_2,y)f_{X_2,Y}(x_2,y)}{f_Y(y)} dx_1dx_2\\ =& \int \int_{\mathbb{R}^2} g_1(x_1)g_2(x_2)\frac{f_{X_1|X_2,Y}(x_1|x_2,y)f_{X_2}(x_2)f_Y(y)}{f_Y(y)} dx_1dx_2\\ \end{aligned} \end{equation}

I'm stuck here.

$\endgroup$
0
$\begingroup$

Is this claim true? Let $X_2,Y$ be independent bernoulli and $X_1$ the sum. $E[X_1X_2\mid Y=1]=2\cdot1/2 + 0\cdot1/2=1$ whereas $E[X_1\mid Y=1]E[X_2]=(2\cdot 1/2 + 1\cdot 1/2)\cdot 1/2=3/4.$

| cite | improve this answer | |
$\endgroup$
  • $\begingroup$ I'm investigating into correctness of the said result, I may have copied it wrongly in class. $\endgroup$ – sixtyTonneAngel Nov 3 '18 at 6:19
0
$\begingroup$

The claim is incorrect. The result requires conditional independence of $X_1$ and $X_2$ given $Y$.

| cite | improve this answer | |
$\endgroup$

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.