I am having a little trouble understanding what exactly is the tail $\sigma$-algebra.

Just so we are all on the same page, my book defined the tail $\sigma$-algebra like this:

Let $X_n$ be a sequence of random variable defined on $(\Omega, \mathcal{A}, P)$. Define $\mathcal{B}_n = \sigma(X_n)$, $\mathcal{C}_n = \sigma(\cup_{p \ge n} \mathcal B_p)$, $\mathcal C_\infty = \cap_{n=1}^\infty \mathcal C_n$. $\mathcal C_\infty$ is called the tail $\sigma$-algebra.

Suppose now that $X_n$ are independent. I am having trouble understanding why the following are true (we are talking about the consequences of Kolmogorov's Zero-One Law): $$\{ \omega: \lim_{n \to \infty} X_n(\omega) \text{ exists} \} \in \mathcal C_\infty$$ $$\limsup_{n \to \infty} X_n \ \ \liminf_{x \to \infty} X_n \ \ \limsup_{n \to \infty} \frac 1n \sum_{p \le n} X_p \ \ \liminf_{n \to \infty} \frac 1n \sum_{p \le n} X_p \text{ are $\mathcal C_\infty$ measurable} $$

I don't have any intuition about why these are true nor what exactly is a tail $\sigma$-algebra apart from the definition, which does not inspire me that much.

Any help is appreciated!

  • $\begingroup$ Which one of these events depend on $X_1$? $\endgroup$ – Did Dec 1 '14 at 23:06
  • $\begingroup$ @Did The first event does not, I guess. The others are r.v. which do depend on $X_1$.. Maybe I am misinterpreting what you said. To be honest I'm a little confused (if that wasn't clear already!) Anyhow I'd appreciate a comment about the downvote and the vote to close, so I can improve something :) $\endgroup$ – Ant Dec 1 '14 at 23:10
  • $\begingroup$ Is $\limsup X_n$ depending on $X_1$? $\endgroup$ – Did Dec 1 '14 at 23:16
  • 1
    $\begingroup$ For example, let $A_k=\{\lim\sum_{n\ge k}X_k\text{ exists}\}$. Then $A_1=A_2=A_k\in\sigma(X_k,X_{k+1},...)$ (because the series converges iff its tail converges). So, $A_1\in\bigcap_{k\ge 1}\sigma(X_k,X_{k+1},...)\equiv \mathcal C_\infty$ $\endgroup$ – d.k.o. Dec 2 '14 at 6:51
  • 1
    $\begingroup$ @d.k.o. I think it is a little clearer now. I would be very grateful if you could make an answer and expand a little over what you just said! :) $\endgroup$ – Ant Dec 4 '14 at 21:42

The answer follows the idea of what @d.k.o. pointed out. For example, for the statement

$$\{ \omega: \exists \lim_{n \to \infty} X_n(\omega) \} \in C_\infty$$

we can argue as follows:

For each n, $X_{m}$ with $m \ge n$ is measurable for the $\sigma$-field ${C}_n = \sigma(\cup_{p \ge n} B_{p})$. Now we use the following fact:

The set of convergence of a sequence of measurable functions is measurable

Therefore the set of convergence of the sequence ${X_{m}}$ with $m \ge n$ is measurable in $C_{n}$. As this set of convergence does not change with n, it follows it belongs to $C_\infty$.

When you understand this proof the rest is easier. Let's proof that:

$\limsup_{n \to \infty} X_n$ is a measurable function in $C_\infty$.

$[\lim sup_{n \to \infty} X_{n} < c]=[\lim sup_{k \to \infty, k \ge n} X_{n} < c]$ which belongs to $C_{n}$ and the statement follows.


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.