0
$\begingroup$

Let $\mathbb{X} := (X_1,X_2)$ be a random variable with given density function $$f_{\mathbb{X}}(x_1,x_2)= \frac{\sqrt{2}}{\pi}\exp\left(-\frac{3}{2}x_1^2-x_1x_2-\frac{3}{2}x_2^2\right),\space \text{for} (x_1,x_2)\in \mathbb{R^2}$$

Show that $f$ is the density of a normal distribution in $\mathbb{R^2}$.

At first I thought the joint density can be expressed as the product of the densities, but completing squares gives me a troublesome expression. I then thought that I can identify it with the bivariate normal distribution: $$p(x_1,x_2)=\frac{1}{2 \pi \sigma_1\sigma_2\sqrt{1-p^2}}\exp\left(-\frac{1}{2}\left(\frac{z}{1-p^2}\right)\right)\,,$$ where $$z=\frac{(x_1-\mu_1)^2}{\sigma_1^2}-\frac{2p(x_1-\mu_1)(x_2-\mu_2)}{\sigma_1\sigma_2}+\frac{(x_2-\mu_2)^2}{\sigma_2^2}$$

If you now take out $-\frac{1}{2}$ in the exponent of $f$ you get $p=-1$, $\mu_1=\mu_2=0$ and $\sigma_1=\sigma_2=\sqrt{\frac{1}{3}}$ and after you multiply $p(x_1,x_2)$ with $\frac{2^{\frac{3}{2}}}{3}$ you exactly have f. So f is normal distributed in $\mathbb{R^2}$. Is this right or one proves this differently?

So $X_1=X_2 \sim N(0,\frac{1}{3})$. Because in the next task one is to calculate the density of $X_1+X_2$ and $X_1-X_2$ and to look whether the are independent or not, but I get the dirac-delta function for the latter one and there I think I might did something wrong. I am thankful for any advice.

$\endgroup$
1
$\begingroup$

HINT 1 : Write the pdf in the form of $c*e^{\frac{-1}{2}}(x'\Lambda x) \quad$ Where $\Lambda$ is symmetric

HINT 2: Now invert $\Lambda$, i.e fine $\Sigma = \Lambda^{-1}$ and then find $A$ such that $AA' = \Sigma$

Then you can argue that if $Z$ follows a standard bivariate normal distribution with mean ${0}$ then $X=AZ$ will also follow bivariate normal distribution with appropriate parameters whose pdf will be the one that you wanted to be proved as a bivariate normal's pdf.

Once you show that $X$ is a bivariate normal, then showing $X_1 + X_2$ to be normal can be done by taking appropriate transformations over $X$, i.e choose $l$ such that $l'X = X_1 + X_2$. In this case $l' = (1 \quad 1)'$. You can find the expressions for its parameters under such transformations in this link under the section "Affine Transformation".

Showing Independence between $X_1 + X_2$ and $X_1 - X_2$ can be done by taking appropriate transformations $BX = (X_1 + X_2, X_1 - X_2)$ and showing that the non diagonal elements are $0$. I.e Covariance between them is $0$ and hence is independent.

Note : Zero Covariance doesn't in general imply Independence but in the case of Bivariate Normal it does.

$\endgroup$

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.