Independent continuous random variables imply joint continuity?

325 Views Asked by At

Suppose I have $\xi_1,\xi_2,\ldots,\xi_n$ real-valued continuous random variables and let $\vec{\xi} = (\xi_1,\xi_2,\ldots,\xi_n)$ a random vector, is it true that if $\xi_i$ are continuous for all $i\in\{1,\ldots,n\}$ and independent, then $\xi_1,\xi_2,\ldots,\xi_n$ are jointly continuous? That is, does the two conditions (I think in this case it would be sufficient but not necessary) imply that $\mathbb{P}_{\vec{\xi}}$ as a push-forward measure is absolutely continuous with respect to the Lebesgue measure on $\mathbb{R}^n$?

The reason I ask this is because in my probability notes, we gave an example of two continuous random variables who aren't jointly continuous. Suppose $\xi\sim \mathcal{U}[0,1]$ (the uniform distribution on $[0,1]$). Let $\eta = \xi$, then $\xi,\eta$ aren't jointly continuous. The proof of this is that we consider the set $C=\{(x,x)\ |\ x\in[0,1]\}$ and consider its push-forward measure and the Lebesgue measures on $\mathbb{R}^2$. It's clear that $m_2(C)=0$ with $m_2$ as the Lebesgue measure on $\mathbb{R}^2$ as $C$ is simply a line. On the other hand, we get $$\mathbb{P}_{(\xi,\eta)}(C) = \mathbb{P}\{0\leq\xi=\eta\leq 1\}= \mathbb{P}\{\omega\in \Omega\ |\ 0\leq \xi(\omega)\leq 1\} = 1.$$ But this construction implies that the two variables are not independent because $\eta=\xi$ is given so they are the same random variable and dependent. On the other hand, if I say we have two random variables, $\xi,\eta$ such that $\xi\sim\mathcal{U}[0,a]$ and $\eta\sim\mathcal{U}[0,b]$ with $a,b>0$ and that they are independent, then I should get that $\xi,\eta$ are jointly continuous. Is that correct?

I suppose my question is that (if my above discussion is correct) is there some generalization of this to the case of all continuous independent random variables?

1

There are 1 best solutions below

6
On BEST ANSWER

The question you are asking is equivalent to the following: If $P_1,P_2,\ldots,P_n$ are proability measures on $\mathbb R$ each of which is absolutely continuous w.r.t. Lebesgue measure $m$ on $\mathbb R$ is it true that the product measure $P_1\times P_2\times\cdots\times P_n$ is absolutely continuous w.r.t. Lebesgue measure $m_n$ on $\mathbb R^n$. The answer is YES. In fact if $\frac {dP_i} {dm}=f_i$ and $f(x_1,x_2,\ldots,x_n)=f_1(x_1)f_2(x_2)\cdots f(x_n)$ then $(P_1\times P_2\times\cdots\times P_n) (E)=\int_E f \, dm_n$. This can be proved easily by first showing it when $E$ is a measurable rectangle (using Fubini's Theorem) and then using standard arguments to show that it holds for all Borel sets $E$.