Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

I need your help with this problem: Suppose $(X, Y)'$ follows a Bivariate Normal Distribution with parameters $μ_1 ,μ_2, σ_1^2, σ_2^2$, and $ρ$. Let $U = X + Y$ and $V = X - Y$. Considering that $X$ and $Y$ are not independent random variables, how will I get the joint distribution of $U$ and $V$. Thanks in advance!

share|cite|improve this question

The vector $(X,Y)$ is gaussian and $(U,V)=(X+Y,X-Y)$ is a linear function of $(X,Y)$ hence $(U,V)$ is gaussian as well. In particular, the distribution of $(U,V)$ is characterized by its mean vector $M$ and covariance matrix $C$. By definition, $$ M=\begin{pmatrix}\mathbb E(U) \\ \mathbb E(V)\end{pmatrix},\qquad C=\begin{pmatrix}\mathrm{var}(U) & \mathrm{cov}(U,V)\\ \mathrm{cov}(U,V)&\mathrm{var}(V) \end{pmatrix}, $$ with

  • $\mathbb E(U)=\mu_X+\mu_Y$, $\mathbb E(V)=\mu_X-\mu_Y$,
  • $\mathrm{var}(U)=\mathrm{var}(X)+\mathrm{var}(Y)+2\mathrm{cov}(X,Y)=\sigma_X^2+\sigma_Y^2+2\varrho$,
  • $\mathrm{var}(V)=\mathrm{var}(X)+\mathrm{var}(Y)-2\mathrm{cov}(X,Y)=\sigma_X^2+\sigma_Y^2-2\varrho$,
  • $\mathrm{cov}(U,V)=\mathrm{var}(X)-\mathrm{var}(Y)=\sigma_X^2-\sigma_Y^2$.

Except when $\varrho^2=\sigma_X^2\sigma_Y^2$, the distribution of $(U,V)$ has a density $f_{U,V}$, defined by $$ f_{U,V}(u,v)=\frac1{2\pi\sqrt{\det C}}\exp\left(-\frac12\left(u-\mathbb E(U),v-\mathbb E(V)\right)^*C^{-1}\left(u-\mathbb E(U),v-\mathbb E(V)\right)\right). $$

share|cite|improve this answer
Is there really no way to express it using the moment-generating function only? We haven't discussed matrices yet. Thanks anyway. – Mach Mar 21 '13 at 0:39
Could you explain express it using the moment-generating function? (What is "it"? Why the MGF, which appears nowhere in the question??) Density of $(X,Y)$ added to the answer. – Did Mar 21 '13 at 7:12

If I'm not mistaken, the general convolution formula yields that the distribution of $U$, $g(u)$, is given by $$g(u) = \int_{-\infty}^\infty \int_{-\infty}^{u-x} \! f(x,y) \, \mathrm{d}y \mathrm{d}x$$ where $f(x,y)$ is the joint pdf of $X$ and $Y$, i.e. your bivariate normal pdf.

Now, since the Normal distribution is stable, you can use this to your advantage and prove that $X+Y$ will be Normal with mean $\mu = \mu_1 + \mu_2$ and variance $\sigma^2 = \sigma_1^2 + \sigma_2^2 + 2 \rho \sigma_1 \sigma_2$. Similarly, get the distribution of $V$.

These two marginals can be "bound" into a joint density by using a copula function.

share|cite|improve this answer
hmmm, but what's needed is the joint of U and V, i.e. the joint of X + Y and X - Y and I'm pretty sure that will not result in a normal distribution. – Mach Mar 20 '13 at 22:15
Ah, yes, missed that part about needing the joint. So you have the two marginals but need to create the joint from these two. Then, you could use a copula function to bind the two marginals. See the Wikipedia page on copulas. – baudolino Mar 20 '13 at 22:42
Okay, thanks! I assume there's really no way to get the joint directly. I've tried using the moment generating function but it ends up in something that's hard to simplify. – Mach Mar 20 '13 at 22:45

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

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