Considering this entry the distribution of the sum of non i.i.d. gaussian variates is also gaussian.

$$ \begin{align*} V = aX + bY &\sim N(a\mu_X + b\mu_Y,\; a^2\sigma_X^2 + b^2\sigma_Y^2 + 2ab\sigma_{X,Y}) \end{align*} $$

What if we have another transformation like W as below:

$$ \begin{align*} W = cX + dY &\sim N(c\mu_X + d\mu_Y,\; c^2\sigma_X^2 + d^2\sigma_Y^2 + 2cd\sigma_{X,Y}) \end{align*} $$

Knowing that X and Y are identically distributed and correlated Gaussian r.v.s what are the marginal distributions of V and W?

Besides, how to get the correlation coefficient of V and W?


1 Answer 1


You have already written out the marginal distributions for $V$ and $W$ above -- they are normal with means and variances as you expressed. To find their correlation, we can write:

$$\text{cov}(V,W) = \text{cov}(aX+bY,cX+dY)$$

which in turn can be decomposed as $$\text{cov}(aX,cX) + \text{cov}(aX,dY) + \text{cov}(bY,cX) + \text{cov}(bY,dY).$$ Notice that $\text{cov}(aX,cX) = ac \cdot \text{var}(X)$ and $\text{cov}(bY,dY) = bd \cdot \text{var}(Y)$. We also know that $\text{cov}(aX,dY) = ad \cdot \text{cov}(X,Y)$ and $\text{cov}(bY,cX) = bc \cdot \text{cov}(X,Y)$. So, you can compute $\text{cov}(V,W)$ by just plugging in these values, which you already know from the joint distribution of $X$ and $Y$.

Finally, to get the correlation coefficient, you can just divide your value of $\text{cov}(V,W)$ by the square root of the variance of $V$ and the square root of the variance of $W$.

  • $\begingroup$ I may get confused about the marginal distribution here. Thanks. $\endgroup$
    – Arjanizary
    Jan 21, 2021 at 19:25

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.