3
$\begingroup$

I am trying to find peer-reviewed references to the following version of the Portmanteau theorem: Let $M$ be a metric space and let $(\mu_n)_{n\in\mathbb N}$ be a sequence of Borel probability measures. Then the following conditions are equivalent:

  • $(\mu_n)_{n\in\mathbb N}$ converges to $\mu$ with respect to the weak$^*$ topology,
  • for every lower semicontinuous function $f\colon M\to\mathbb R$ bounded from below one has $$\liminf_{n\to\infty}\int f\text{d}\mu_n\geq \int f\text{d}\mu.$$
$\endgroup$
4
  • 1
    $\begingroup$ These days I'd expect to find it in books instead of articles. Have you checked in Billingsley? $\endgroup$ Apr 20, 2018 at 20:10
  • $\begingroup$ You're welcome; was it there? If you find a satisfactory reference, you might like to post your own answer to this question. $\endgroup$ Apr 21, 2018 at 20:38
  • $\begingroup$ The proof is about two lines, so if I were using this in a paper, I probably wouldn't bother with a reference - maybe just include a brief hint at the proof. Forward: consider a sequence of bounded continuous functions increasing to $f$. Reverse: what user111 said, or: if $f$ is bounded continuous, apply the second statement to $f$ and $-f$. $\endgroup$ Apr 22, 2018 at 3:58
  • 1
    $\begingroup$ I know the proof, but I would prefer to cite it. I think the reference posted below by user111 is very good. Anyway, thank you for your help! $\endgroup$ Apr 22, 2018 at 16:37

1 Answer 1

3
$\begingroup$

The Portmanteau theorem does not seem to be stated in this form in Billingsley or other classical references that I checked. A possible reference for the direct implication is Theorem A.3.12. p.378 of

Dupuis, P., Ellis, R.S., A weak convergence approach to the theory of large deviations. Wiley Series in Probability and Statistics, Wiley-Interscience, New York, 1997

The second statement implies that for an open set $O$ of $M$, $$\liminf_{n\to\infty}\mu_n(O)\geq\mu(O),$$ so the reverse implication just follows from the above inequality since it is one of the equivalent properties to weak-* convergence, as stated in the usual version of the Portmanteau theorem.

$\endgroup$
0

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.