2
$\begingroup$

Say I have the following : $$ (X, Y) \sim N_2(\mu, \Sigma) $$

Then what would be the distribution of $(2X,2Y)$ ?

Let $\Sigma = \begin{pmatrix} \sigma_1^2 & \rho\sigma_1\sigma_2\\ \rho\sigma_1\sigma_2 & \sigma_2^2 \end{pmatrix}$

I think $\rho$ shouldn't change since : $$ corr(π‘Žπ‘‹+𝑏,π‘π‘Œ+𝑑)=corr(𝑋,π‘Œ) $$ and that I should use the univariate transformation for the sigmas, so it would be : $$ (2X, 2Y) \sim N_2(2\mu, 4\Sigma) $$ but can't prove it, and my simulations don't seem to back it. Does anybody have a clue about this ? I can't find anything on the internet about multiplying a bivariate gaussian random variable with a constant

$\endgroup$
1
  • $\begingroup$ The random variable setting is a red herring: this is just a change-of-units problem, no different than converting pecks to gallons. When you view it this way you should immediately be able to write down the answer. $\endgroup$
    – whuber
    Commented Feb 12, 2020 at 15:21

1 Answer 1

1
$\begingroup$

First, they'll be still jointly normal because this is a linear transformation. The rest is pretty straightforward. Each entry in the covariance matrix and the mean vector can be calculated individually, e.g. $$\operatorname{var}(2X)=4\operatorname{var}(X), \operatorname{cov}(2X,2Y)=4\operatorname{cov}(X,Y),E[2X]=2E[X]$$ So, your result is correct.

$\endgroup$

Your Answer

By clicking β€œPost 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.