\subsubsection*{Joint Distributions} Given two random variables $X$ and $Y$, a {\em joint random variable\/} is defined to be a random variable whose type is ordered pair, and whose restriction to the first member of the pair is $X$, and whose restriction to the second member of the pair is $Y$. In real-life applications, any two random variables have a naturally given joint random variable. Consider the roulette example where you and your friend bet on different numbers on the same spin of the wheel. Recall that \begin{itemize} \item $X = 35$ with probability $\frac{1}{38}$ \item $X = -1$ with probability $\frac{37}{38}$ \end{itemize} while $Y$ has the same values with the same probabilities. (We say that $X$ and $Y$ are ``identically distributed.") The joint random variable is then the ordered pair $(X,Y)$, distributed as follows: \begin{itemize} \item $(X,Y) = (35,35)$ with probability 0. \item $(X,Y) = (35,-1)$ with probability $\frac{1}{38}$. \item $(X,Y) = (-1,35)$ with probability $\frac{1}{38}$. \item $(X,Y) = (-1,-1)$ with probability $\frac{18}{19}$. \end{itemize} Given two distributions $\lambda$ and $\mu$, a {\em joint distribution\/} of $\lambda$ and $\mu$ is defined to be the distribution of the joint random variable of any two random variables whose distributions are $\lambda$ and $\mu$. Note that the joint distribution of $\lambda$ and $\mu$ is not uniquely defined, as we can see by considering the roulette example. Suppose that $\lambda_1 = \mu_1 = \frac{1}{38}$, and that $\lambda_2 = \mu_2 = \frac{37}{38}$, and $\nu$ is a joint distribution of $\lambda$ and $\mu$. Then $\nu_{(1,1)}$ could be any real number between 0 and $\frac{1}{38}$. In fact, $\nu$ is required only to satisfy the following conditions: \begin{itemize} \item $\nu_{(i,j)}\ge 0$ for each $i,j$. \item $\nu_{(1,1)} +\nu_{(1,2)} =\lambda_1 = \frac{1}{38}$ \item $\nu_{(2,1)} +\nu_{(2,2)} =\lambda_2 = \frac{37}{38}$ \item $\nu_{(1,1)} +\nu_{(2,1)} =\mu_1 = \frac{1}{38}$ \item $\nu_{(1,2)} +\nu_{(2,2)} =\mu_2 = \frac{37}{38}$ \end{itemize} There are infinitely many different choices of $\nu$ that satisfy those conditions. \begin{exercise}{\rm Find the three different joint distributions $\nu$ for the roulette example, corresponding to the following three situations: \begin{itemize} \item You and your friend bet on different numbers on the same spin of the wheel. \item You and your friend bet on different numbers on the different spins of the wheel. \item You and your friend bet on the same number on the same spin of the wheel. \end{itemize} } \end{exercise}