$\mathbf{Y}=(Y_1, Y_2)^T$ is a bivariate Gaussian random vector with mean zero.

I want to prove that $Cov[Y_1^2, Y_2^2]=2\times(Cov[Y_1, Y_2])^2$.

I know that

$$Cov[Y_1^2, Y_2^2]=E[Y_1^2Y_2^2]-E[Y_1^2]E[Y_2^2]=E[Y_1^2Y_2^2]-Var[Y_1]Var[Y_2]$$


$$2(Cov[Y_1, Y_2])^2=2(E[Y_1Y_2]-E[Y_1]E[Y_2])^2=2(E[Y_1Y_2])^2$$ as $\mu_{Y_1}$,$\mu_{Y_2}$ are 0.

What is the next step I should proceed?


Some facts:

  1. If $(Y_1,Y_2)$ is jointly normal (with mean 0) with parameters, $\rho, \sigma_1$ and $\sigma_2$, then $$Y_1 | Y_2 \sim N\left(\frac{\rho\sigma_1}{\sigma_2}Y_2, (1-\rho^2)\sigma_1^2\right)$$

  2. The fourth central moment of a normal random variable is $\mathbb{E}[Y] = 3\sigma^4$.

Using these facts, we can write,

$$\begin{eqnarray} \mathbb{E}[Y_1^2Y_2^2] &=& \mathbb{E}[\mathbb{E}[Y_1^2Y_2^2 | Y_2]]\\ &=& \mathbb{E}[Y_2^2\mathbb{E}[Y_1^2|Y_2]]\\ &=& (1-\rho^2)\sigma_1^2\mathbb{E}[Y_2^2] + \frac{\rho^2\sigma_1^2}{\sigma_2^2}\mathbb{E}[Y_2^4]\\ &=& (1-\rho^2)\sigma_1^2\sigma_2^2 + 3\rho^2\sigma_1^2\sigma_2^2 \end{eqnarray}$$

Substitute this into your expression for $Cov(Y_1^2,Y_2^2)$ to get $2\rho^2\sigma_1^2\sigma_2^2$ which is identically $2(Cov(Y_1,Y_2))^2$.


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.