0
$\begingroup$

There are many materials about the reverse question: "Does uncorrelatedness tell us something about independence?" But how to answer the question I've posed and why? Is there some simple counterexample?

$\endgroup$
0
$\begingroup$

I will try to give an intuitive answer.

To say that two random variables $X$ and $Y$ are uncorrelated means that $\mathbb{E}[XY] = \mathbb{E}[X] \mathbb{E}[Y]$. On the other hand, saying $X$ and $Y$ are independent means that their joint distribution is the product of their marginal distributions, $p(X,Y) = p(X) \cdot p(Y)$; or, equivalently, that either joint distribution is the same as its marginal distribution $p(Y|X) = p(Y)$.

Now, suppose that $X$ and $Y$ are correlated. Then there must be values $x_{1}$ and $x_{2}$ for which $\mathbb{E}(Y|X=x_1) \neq \mathbb{E}(Y|X=x_2)$.*

Since the expectations are unequal, it is clear that $p(Y|X=x_1) \neq p(Y|X=x_2)$. But if $X$ and $Y$ were independent, we would have $p(Y|X=x_1) = p(Y) = p(Y|X=x_2)$. Thus correlation implies dependence and, by the contrapositive, independence implies uncorrelation.

* This is easiest to see by assuming that $\mathbb{E}[Y]$ is constant and working out $\mathbb{E}[XY]$ by integration

$\endgroup$

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.