So, I recently (re-)discovered that random variables learned in elementary probability such as the exponentially distributed random variable $X$ with cdf $F_X(x) = 1-e^{- \lambda x}$ can be explicitly represented as

$$X(\omega) := \frac{1}{\lambda} \ln(\frac{1}{1-\omega})$$

where the probability space is $((0,1), \mathscr B(0,1), \mu)$ where $\mu$ is Lebesgue measure.

This can be derived with the formula

$$X(\omega) = \sup\{x \in \mathbb{R}: F_X(x) < \omega\}$$

This is apparently called Skorokhod representation (so-called in David Williams' Probability with Martingales).

Now, apparently, $X(1-\omega)$ is not only also exponential but also does it have the same parameter $\lambda$ as $X(\omega)$, i.e. they have the same distribution.

Similarly, for $X \sim Be(p)$, I have found the following to be $\sim Be(p)$:

  1. $1_{(0,1-p)}(\omega)$

  2. $1_{(0,1-p)}(1-\omega) = 1_{(p,1)}(\omega)$

  3. $1_{(0,1-p)}(1-(\omega+\frac{p}{2})) = 1_{(p,1)}(\omega+\frac{p}{2}) = 1_{(\frac{p}{2},1-\frac{p}{2})}(\omega)$

  4. $1_{(0,1-p)}(1-(\omega+\varepsilon)) = 1_{(p,1)}(\omega+\varepsilon) = 1_{(\varepsilon,1-p+\varepsilon)}(\omega)$ for $0< \varepsilon < p < 1$

So, for what conditions on $g(\omega)$ are sufficient or necessary s.t. $X(g(\omega))$ has the same distribution as $X(\omega)$? I guess we need $g(\omega) : (0,1) \to (0,1)$.

Also does $g(\omega) = 1-\omega$ work for any X? It seems so:

$$X(1 - \omega) := \sup\{x \in \mathbb R | F_X(x) < 1 - \omega\}$$

$$ = \sup\{x \in \mathbb R | \omega \le 1 - F_X(x)\}$$

$$ = \sup\{x \in \mathbb R | \omega < 1 - F_X(x)\}$$

$$ = \sup\{x \in \mathbb R | \omega \ge F_X(x)\}$$

$$ = \sup\{x \in \mathbb R | \omega > F_X(x)\} := X(\omega)$$


Basically $\omega$ has the same probability of being in $[F_X(x),1)$ as $(0,F_X(x)]$ because both intervals have the same Lebesgue measure.


$X(g(\omega))$ has the same distribution as $x$ if $g:(0,1)\to (0,1)$ is what is called a measure preserving function. This means $m(g^{-1}A))=m(A)$ for all Borel sets $A$ (where $m$ is the Lebesgue measure on $(0,1)$.

BCLC edit:

Proposition: If $g:(0,1)\to (0,1)$ is a measure preserving function, i.e. $Leb(g \in B) = Leb(B) \forall B \in \mathscr B(\mathbb R)$ then $F_{X(g)} = F_X$ or equivalently $\mathcal L_X(B)=\mathcal L_{X(g)}(B)$.


$$P(X(g(\omega)) \in B)$$

$$ = P(g(\omega) \in X^{-1}(B))$$

$$ = P(\omega \in g^{-1}(X^{-1}(B)))$$

$$ = P(\omega \in X^{-1}(B)) \tag{*}$$

$$ = P(X(\omega) \in B)$$


$(*)$ Here we make use that $P$ is Leb on $((0,1), \mathscr B(0,1))$.

  • 2
    $\begingroup$ Examples of measure preserving maps can be found books on Ergodic Theory $\endgroup$ Apr 29 '18 at 12:40
  • $\begingroup$ 1. Proof please? 2. Any relation of my question or your answer to this mathoverflow.net/q/227075/69696? $\endgroup$
    – BCLC
    Apr 29 '18 at 13:53
  • $\begingroup$ Well after proving for myself, I think I kind of see the relation. Thanks Kavi Rama Murthy! $\endgroup$
    – BCLC
    May 3 '18 at 4:37

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.