Let $X_1,X_2,\dots$ be a sequence of real-valued random variables where $X_n:\Omega \to \mathbb{R}$.

For any sequence of events $A_n \subset \Omega$ define $\limsup_nA_n := \bigcap_{n=1}^{\infty}\bigcup_{m=n}^{\infty}A_m$.

Moreover, define the (extended-real) random variable $Y \equiv (\limsup_nX_n)$ by $Y(\omega) := \limsup_n\{X_n(\omega) \mid n=1,2,\dots\}$.

How would you establish that

$$ \limsup_{n}\{\omega \mid X_n(\omega) \in B\} = \{\omega \mid (\limsup_nX_n)(\omega)\ \in B \}, \qquad (*) $$

for any Borel set $B$? Is the proposition $(*)$ even true?


After further thought I believe that $(*)$ is in general false.

Let $B = \{ b\}$ be a singleton set. Observe that

  1. $\omega \in \{ \omega' \mid (\limsup_nX_n)(\omega') = b \}$ if and only if $\limsup\{X_1(\omega), X_2(\omega), \dots \} = b$.

  2. $\omega \in \limsup_n\{ \omega' \mid X_n(\omega') = b \}$ if and only if for all $n$ there exists $m>n$ such that $X_m(\omega)=b$, or in other words $X_n(\omega) = b$ infinitely often.

Let $\omega$ be such that the sequence $(X_n(\omega))_{n=1}^{\infty} = (0,1,0,1,0,1,0,1,\dots)$ and put $b=0$. Then $(\limsup_nX_n)(\omega) = 1$, however $X_n(\omega) = 0$ occurs infinitely often.


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.