1
$\begingroup$

I'm reading David Williams' Probability with Martingales.

On page 41 he performs the following manipulation

$$ L := \lim \sup(X_n/\log n)\\ \implies \mathbb{P}(L \geq 1) \geq \mathbb{P}(X_n > \log n,\text{ infinitely often}) $$

It appears to me as if by some sleight of hand $\lim \sup$ is being used in two very different ways here. The first $\lim \sup$ is one of a function, presumably pointwise. The second $\lim \sup$, which appears in the form of "infinitely often" is one of sets.
Fundamentally, what I think it boils down to is how the following two example sets, written out in all detail, relate to another.

$$ \left\{\omega : \lim_{m \rightarrow \infty} \sup_{n \geq m} X_n(\omega) \leq x\right\} $$

$$ \bigcap_m \bigcup_{n \geq m} \{\omega : X_n(\omega) \leq x\} $$

I can sorta see that the first is contained in the second. But I'm not convinced that the second is contained in the first.

$\endgroup$
0
$\begingroup$

The implication has a $\geq$ in it. If $X_n>\log n$ infinitely often for a fixed $\omega$, that means there's a sequence $n_i$ with $X_{n_i}(\omega)>\log n_i$. This implies $\limsup_n X_n(\omega)/\log_{n_i}>1$ by definition of limsup. Hence the $\geq$ follows.

$\endgroup$
0
$\begingroup$

Alternatively, think of an event (such as $E_n=[X_n>\log n]$ as a 0-1 valued function, the indicator of the event: $\mathbf 1_{E_n}(\omega)=1 \iff X_n(\omega)>\log n$, and so on. The definition of $\limsup$ for events was chosen so that this kind of punning works out correctly.

$\endgroup$

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.