Does $E[X\vert Y=y] = \gamma y$ imply that $X,Y$ are jointly normal?

31 Views Asked by At

I know that if $X,Y$ are jointly normal with unit variances, then $E[X\vert Y=y] = \rho y$, which is of the form in the title (with $\rho$ being the covariance). Also, I think as long as the means are $0$ then the conditional expectation will have the proper form.

However, what I don't know is whether the other direction is true. That is, I don't know whether $E[X\vert Y=y] =\gamma y \implies $ that $X,Y$ are jointly normal.

I feel like this isn't true, but I can't prove that it isn't (I don't even know how I'd approach it).

I'm not looking for a proof, although I'm okay if one is posted.

1

There are 1 best solutions below

0
On BEST ANSWER

Answer is no.

Carmichael561 gave a simple counterexample.

"If $X=\gamma Y$ then $E[X\vert Y=y] = \gamma y$, regardless of the distribution of $Y$"