3
$\begingroup$

The exercise is the following. Let $(X,Y)$ be a bi-variate normal vector with $\mu=(2,3)$ and covariance matrix $$\Sigma=\begin{pmatrix}1 & -2 \\ -2 & 4\end{pmatrix}.$$ What is the distribution of the random variable $X+Y$?

My try is: Consider the moment generating functions of $X$ and $Y$. Since $\sigma_X^2=1$, $\sigma_Y^2=4$, $\mu_X=2$ and $\mu_Y=3$, we have $$\varphi_X(t)=\exp(2t+\frac{t^2}{2})$$ $$\varphi_Y(t)=\exp(3t+2t^2).$$

Thus, $$\varphi_{X+Y}(t)=\varphi_X(t)\varphi_Y(t)=exp(5t+\frac{5t^2}{2}).$$

Hence, $X+Y\sim N(5,5)$ since the moment generating function of a normal r.v. is given by $\varphi_X(t)=\exp(\mu_X t + \frac{\sigma_X^2 t^2}{2})$.

I am asking this because my teacher put as a solution $X+Y\sim N(5,1)$. Any feedback?

EDIT:

I think that we can not use $\varphi_{X+Y}(t)=\varphi_X(t)\varphi_Y(t)$ since $X$, $Y$ are not independent.

$\endgroup$
2
$\begingroup$

The variance of $X+Y$ is $$ \operatorname{Var}(X)+\operatorname{Var}(Y)+2\operatorname{Cov}(X,Y)=1+4+2\times(-2)=1. $$ So $N(5,1)$ is correct. Your problem is the assertion $\varphi_{X+Y}(t)=\varphi_X(t)\varphi_Y(t)$ which doesn't work in this case because $X$ and $Y$ are dependent.

$\endgroup$
2
$\begingroup$

The two marginal distributions are clearly non independent which means you have to find the sum of two non i.i.d. gaussians.

$X+Y \sim N(\mu_x + \mu_y, \sigma_x^2 + \sigma_y^2 +2 \sigma_{x,y})$

$\endgroup$

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.