7
$\begingroup$

I'm looking for a reference book or article for the following two facts. In both statements, a Polish space $E$ and an ambient probability space $(\Omega, {\cal A}, \Pr)$ are given, and I consider the topology of weak convergence on the space $E'$ of probability measures on $E$ (thus $E'$ is Polish too).

  • 1) Let $X$ be a random variable taking its values in $E$, and let ${\cal B} \subset {\cal A}$ be a $\sigma$-field. Then the conditional law ${\cal L}(X \mid {\cal B})$ is a $E'$-valued random variable. This should be a consequence of the measurability of the conditional expectations $\Bbb E[f(X) \mid {\cal B}]={\cal L}(X \mid {\cal B})(f)$ but I don't find any standard probability theory book asserting this fact (I'd prefer a "standard" probability theory book rather than a more technical book about random measures).

  • 2) Let $(\mu_n)$ be a sequence of random probability measures on $E$ (in other words the $\mu_n$ are $E'$-valued random variables). Let $\mu_\infty$ be another random probability on $E$ and assume that for each bounded continuous function $f\colon E\to \Bbb R$, the convergence $\mu_n(f) \to \mu_\infty(f)$ holds almost surely. It is tempting to conclude that $\mu_n \to \mu_\infty$ but this is not straightforward since the full set of convergence in $\mu_n(f) \to \mu_\infty(f)$ could depend on $f$. So I'm looking for a reference book showing this almost sure convergence in $E'$. Theorem 7.5.2 in this book by Kuksin's and Shirikyan answers the question but it is stronger than desired because it does not assume the presence of $\mu_\infty$. Moreover I'd prefer a more standard probability/measure theory book.

Actually I would be satisfied to find a reference for the special case when $E$ is compact. Thanks in advance.

Improve this question
$\endgroup$
3
  • 2
    $\begingroup$ I think a relatively readable book on these issues is "Random Probability Measures on Polish Spaces" by Crauel. $\endgroup$
    – Michael Greinecker
    Jun 22, 2013 at 10:57
  • 2
    $\begingroup$ For (2), in the case that $E$ is compact, you can work on a countable dense subset of $C(E)$. And in the locally compact case, you get weak convergence as soon as you have $\mu_n(f) \to \mu_\infty(f)$ for all $f \in C_0(E)$, or a countable dense subset thereof. $\endgroup$
    – Nate Eldredge
    Jun 22, 2013 at 11:22
  • $\begingroup$ Thanks. Here is an interesting reference given in Kuksin & Shirikyan's book tandfonline.com/doi/abs/10.1080/17442500600745359#.UcXQjPnTquk $\endgroup$
    – Stéphane Laurent
    Jun 22, 2013 at 16:48

1 Answer 1

Reset to default
2
$\begingroup$

For the first question I read a paper today related to Markov Kernels which state it in a more general case (for Markov Kernels)

It has no demonstration at all, but it is a good reference for this basic facts and definitions.

Check it out in section 2.

http://www.sciencedirect.com/science/article/pii/S0167715212004579 The Author is A.G. Nogales, "The existence of regular conditional Probabilities for Markov Kernels".

Improve this answer
$\endgroup$

Your Answer

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

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