Let F be a distribution function. On $(\Omega, \mathfrak{F}, P)=((0,1), \mathfrak{B}(0,1),\lambda)$ where $\lambda$ denotes Lebesgue measure.

Define X: $\Omega \to \mathbb{R}$ by $X(\omega) = \sup(y \in \mathbb{R}: F(y) < \omega)$.

1 Show that $\forall x \in \mathbb{R}, (\omega: X(\omega) \leq x) = (\omega: \omega \leq F(x))$

2 Show that X is a RV in $(\Omega, \mathfrak{F}, P)$ and that $F_X = F$.

1 LHS = $(\omega: X(\omega) \leq x)$

= $(\omega: X(\omega) \in (-\infty, x])$

= $(\omega: \sup(y \in \mathbb{R}: F(y) < \omega) \in (\infty, x])$

RHS = $(\omega: \omega \leq F(x))$

= $(\omega: \omega \in (-\infty, F(x)])$

= $(\omega: \omega \in (-\infty, P(X^{-1}( \ (-\infty,x] \ ))))$

= $(\omega: \omega \in (-\infty, P(LHS)))$

I'm stuck. Help please? :(

Cross posted : https://stats.stackexchange.com/questions/110704/let-f-be-a-distribution-function-prove-that-x-is-a-rv

  • $\begingroup$ A distribution function is in principle a function on $\mathbb R$, or more general on $\mathbb R^n$. A measurable function $X:\Omega\rightarrow\mathbb R$ will induce a distribution function $F_X$. $\endgroup$ – drhab Aug 5 '14 at 12:21
  • $\begingroup$ You should start here just with: 'Let $F$ be a distribution function.' Not a distribution function on $(\Omega,...)$. $\endgroup$ – drhab Aug 5 '14 at 12:33
  • $\begingroup$ @drhab 1 Edited 3rd statement. sorry for confusion. 2 Do you mean that the existence of a distribution function implies the existence of a random variable inducing it? I think this is precisely what is being attempted to be shown. $\endgroup$ – BCLC Aug 5 '14 at 12:34
  • $\begingroup$ @drhab I edited exactly as it was given to me. "Let F...function. On...measure." The second statement "On...measure." is a fragment. $\endgroup$ – BCLC Aug 5 '14 at 12:35
  • $\begingroup$ To be shown is: if $F$ is some distribution function, then we can always construct some rv $X$ such that its distribution function $F_X$ will satisfy $F_X=F$. A start has been made with this construction and you are asked to prove that the constructed $X$ here satisfies this condition. $\endgroup$ – drhab Aug 5 '14 at 12:36

If $X\left(\omega\right)\leq x<z$ then if follows immediately from the definition $X\left(\omega\right):=\sup\left\{ y\mid F\left(y\right)<\omega\right\} $ that $F\left(z\right)\geq\omega$.

The fact that $F$ is continuous on the right then allows the conclusion that also $F\left(x\right)=\lim_{z\rightarrow x+}F\left(z\right)\geq\omega$.

In conversely $F\left(x\right)\geq\omega$ then $X\left(\omega\right)=\sup\left\{ y\mid F\left(y\right)<\omega\right\} \leq x$ because $F$ is non-decreasing.

Proved is now that $$X\left(\omega\right)\leq x\iff\omega\leq F\left(x\right)$$

This equation gives us the second part:

$X:\Omega\rightarrow\mathbb{R}$ is a measurable function and this with: $$F_{X}\left(x\right)=P\left\{ \omega\mid X\left(\omega\right)\leq x\right\} =\lambda\left\{ \omega\in\left(0,1\right)\mid\omega\leq F\left(x\right)\right\} =\lambda\left((0,F\left(x\right)]\right)=F\left(x\right)$$

  • $\begingroup$ THANK YOU SO MUCH. Question on last line: Why is it $F(x)]$ and not $F(x))$ ? I was thinking that there was a possibility that $F(x)=1$ so we would have to exclude it. I know it does not matter for Lebesgue measure so the answer is the same regardless. I was just thinking that the LHS of the last inequality should be piecewise or something: ] if $F(x) < 1$ and ) if $F(x) \leq 1$ $\endgroup$ – BCLC Aug 6 '14 at 4:22
  • 1
    $\begingroup$ Mwah.. I think that $\Omega=\left(0,1\right)$ is a somewhat stupid choice here. To avoid confusion on this it is better to choose for $\Omega=\left[0,1\right]$. Then everythings goes fine too and issues like you mention do not arise. $\endgroup$ – drhab Aug 6 '14 at 6:59
  • $\begingroup$ Thanks. I'll ask my prof. Fingers crossed figuring that out isn't part of the homework. If it is though...piecewise then? $\endgroup$ – BCLC Aug 6 '14 at 9:29
  • 1
    $\begingroup$ If you wish, but I wouldn't choose for that. Elegance is lost. You can also it do with a last line that simply leaves out $=\lambda\left((0,F\left(x\right)]\right)$. Then the disputable interval $(0,F\left(x\right)]$ is kept out of sight. $\endgroup$ – drhab Aug 6 '14 at 9:43

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.