Is $E(E(X|Y))^2=E(E(X|Y)X)$?

75 Views Asked by At

Is the equality $E(E(X|Y))^2=E(E(X|Y)X)$ true or does there need to be some relationship between the random variables $X$ and $Y$ such that equality holds.

Working with the left hand side, we have \begin{align} E(E(X|Y)X)&=E(E(X|Y)E(E(X|Y)))\\ &=E(E(X|Y)E(X|Y))\\ &=E(E(X|Y))^2 \end{align} where we take out the first expectation in $E(E(X|Y))$ in the first line because $E(\text{constant})=\text{constant}$. I am merely not convinced of this result.

I validated this equality with a simple example, let $Y=1$ if a fair coin turns up heads and 0 if tails. If $Y=1$, toss a die and let $X$ be the value shown on the face of the die. If $Y=0$, toss the coin again and let $Y$ be 1 if the fair coin is heads and 0 if tails. It seems that in this example, both sides are $\frac{50}{8}$ if I recall correctly.

Clearly, $X$ and $Y$ are not independent since $Y$ determines whether a coin or a die is thrown $(X)$. So, it does not seem like we need to use independence, which may indicate that there is no problem with the steps we followed above. Is this true?

2

There are 2 best solutions below

3
On BEST ANSWER

In general for suitable Borel-measurable $f:\mathbb R\to\mathbb R$: $$\mathbb E[f(Y)X]=\mathbb E[\mathbb E(f(Y)X\mid Y)]=\mathbb E[f(Y)\mathbb E(X\mid Y)]$$

In the special case where $f(Y)=\mathbb E(X\mid Y)$ this leads to:$$\mathbb E[\mathbb E(X\mid Y)X]=\mathbb E[\mathbb E(X\mid Y)\mathbb E(X\mid Y)]=\mathbb E[\mathbb E(X\mid Y)^2]$$

However be aware of the fact that in general $\mathbb E[\mathbb E(X\mid Y)^2]\neq\mathbb E[\mathbb E(X\mid Y)]^2$ so the last equality in your question is not okay.

2
On

The equation is valid and it is proved by starting with the right side and conditioning on $Y$. The conditional expectation of the right side w.r.t. Y is $(E(X|Y))^{2}$ so the rigth side equals the expectation of this (which is the left side).