1
$\begingroup$

Let $F:\mathbb{R}\rightarrow \mathbb{R}$ be a distribution function (CDF)

In this case, we can define the inverse $X$ of $F$, and it is a random variable on $(0,1)$ such that $F_X=F$.

Hence, every distribution (CDFs) can be viewed as the cdf of a random variable on $(0,1)$.

Is there an analogous result for joint distribution functions (CDFs)?

That is, for a fixed $n$, does there exists a probability space $(\Omega,\mathscr{F},P)$ such that every joint distribution function $F:\mathbb{R}^n\rightarrow \mathbb{R}$ is $F_X$ for some $n$-dimensional random vector $X$ on $(\Omega,\mathscr{F},P)$?

$\endgroup$
  • $\begingroup$ It is not true that every CDF can be viewed as the CDF of some random variable on $(0,1)$. Perhaps you are referring to the fact that if $U$ is a Uniform random variable on $(0,1)$, then $F^{-1}(U)$ is a random variable whose CDF is $F$ (where $F^{-1}$ is the generalized inverse function, link.) Your final question's answer is true, every joint CDF is the CDF of some $n$-dimensional random vector. $\endgroup$ – Milind Apr 19 '18 at 15:55
  • $\begingroup$ @Milind It is true. For example, see the proof for theorem 1.2.2 in durrett - probability and example $\endgroup$ – Rubertos Apr 19 '18 at 16:08
  • $\begingroup$ @Milind and for the second question, I am not asking that. I know that. I am asking wheter there exists a “universal” probability space $\Omega$ that every joint distribution of a random vecot $X$ becomes a joint distribution of a random vector $X’$ on $\Omega$. $\endgroup$ – Rubertos Apr 19 '18 at 16:09
  • $\begingroup$ Ah ok, Durrett 1.2.2 says that there exists a random variable defined on the probability space $(0,1)$ with Lebesgue measure and Borel measurable sets. So it is a difference of terminology; the phrase "$X$ is a random variable on the space $S$" usually is taken to mean $X$ takes values in $S$, not that $S$ is the probability space. $\endgroup$ – Milind Apr 19 '18 at 20:11
1
$\begingroup$

The answer to your question is yes; As in the 1-D case, you can take the probability space $([0,1], \mathcal F, \mathcal L)$, where $\mathcal F$ is the Borel sets and $\mathcal L$ is the Lebesgue measure, and for any CDF $F$ there exists $X: [0,1]\to\mathbb R^n$ such that $X$ has CDF $F$.

This is an instance of the more general fact that any Borel probability measure on a complete separable metric space is a pushforward of Lebesgue measure on $[0,1]$. See chapter 13 of Dudley's Real Analysis and Probability.

$\endgroup$
  • $\begingroup$ Thank you, but I cannot find this content in Dudley’s book.. Could you tell me which page it is? $\endgroup$ – Rubertos Apr 19 '18 at 21:42

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.