3
$\begingroup$

Suppose I have a measurable space $(\Omega, \Sigma)$ and a function $f: \mathbb{R} \times \mathbb{\Sigma} \rightarrow [0,1]$ such that for any $x \in \mathbb{R}$ the tuple $(\Omega, \Sigma, f(x, \_))$ is a measure space.

Is it the case that $$\int_{\mathbb{R}} \left[ \int_{\Omega} df(x, \_) \right] dx = \int_{\Omega} d \left[ \int_{\mathbb{R}} f(x, \_) dx \right] \ ?$$

I suspect that this comes down to the question of whether we can interchange the Radon-Nikodym derivative and integration operations, but I am not sure. If this question doesn't make sense because of a missing assumption please let me know :)

Improve this question
$\endgroup$
3
  • 4
    $\begingroup$ This just might just be my general ignorance, but I'm struggling a bit to make sense of your question. It seems to me like you have a family $(\nu_t)_{t\in \mathbb{R}}$ of measures and want to know whether/when it's possible to make sense of the 'integral' $ \nu = \int_\mathbb{R}\nu_t\hspace{.2pc}\mathrm{d}t$ and, if so, whether $ \nu(E) = \int_\mathbb{R} \nu_t(E)\hspace{.2pc} \mathrm{d}t$ for $E\in \Sigma$. Is that anywhere close? $\endgroup$
    – DCM
    Dec 31, 2021 at 21:06
  • 3
    $\begingroup$ (my $\nu_t$ is your $f(t,\cdot)$ by the way - I know I should really stick with your notation, but writing it in a more familiar way might help people (me) recognise certain things more easily) $\endgroup$
    – DCM
    Dec 31, 2021 at 21:07
  • $\begingroup$ @DCM yes I believe that is right $\endgroup$
    – gigalord
    Jan 1, 2022 at 16:16

1 Answer 1

Reset to default
3
$\begingroup$

Provided $x\mapsto f(x,B)$ is Borel measurable for each $B\in\Sigma$, you can define $$ \mu(B):=\int_{\Bbb R} f(x,B) dx, $$ and check (using Tonelli's theorem) that $\mu$ is a measure. If $H\ge 0$ is a measurable function on $\Omega$, then $$ \int_\Omega H d\mu=\int_{\Bbb R}\left[\int_\Omega H(\omega) d_\omega f(x,\omega)\right] dx, $$ by the usual simple function approximation.

Improve this answer
$\endgroup$
5
  • $\begingroup$ The OP's notations are somewhat poorly chosen, leading one to commit the error that you yourself make above: the second argument of $f$ is not a point $\omega \in \Omega$, but a measurable subset $B \in \Sigma$, rendering the inner integral on the right side meaningless. $\endgroup$
    – Alex M.
    Jan 1, 2022 at 18:04
  • $\begingroup$ No. My notation is also poorly chosen! The inner integral on the right is intended to be the integral of $H$ with respect to the measure $f(x,\cdot)$. Perhaps I should have used the probabilist's notation $\int_\Omega H(\omega) f(x,d\omega)$. $\endgroup$
    – John Dawkins
    Jan 1, 2022 at 18:15
  • 1
    $\begingroup$ Oh, then it's probably $\int _\Omega H \, \mathrm d [f(x, \cdot)]$. $\endgroup$
    – Alex M.
    Jan 1, 2022 at 18:18
  • $\begingroup$ Thank you! I am not sure I follow the last step though. What do you mean by "usual simple function approximation"? $\endgroup$
    – gigalord
    Jan 2, 2022 at 14:06
  • $\begingroup$ The equality is shown for $H=1_B$. By linearity it extends to non-neg simple functions (positive linear combinations of indicators). Arbitrary $H\ge 0$ is the incr. limit of a sequence of simple functions. $\endgroup$
    – John Dawkins
    Jan 2, 2022 at 17:07

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.