I am studying probability theory myself, so I have been asking questions a lot recently. Please help.

This question comes from Rosenthal's 3.6.13

Let $X_1, X_2,\dots$ be defined jointly on some probability space $(\Omega, \mathcal{F}, \mathbb{P})$, with $\sum_{i=1}^\infty i^2\mathbb{P}(i\leq X_n<i+1)\leq C\leq \infty$ for all $n$. Prove that $\mathbb{P}(X_n \geq n\ i.o.) = 0$

I am thinking if I can prove $\sum_{n=1}^\infty \mathbb{P}(X_n \geq n) < \infty$, then Borel-Cantelli Lemma can apply, but got no luck. What I got are:

\begin{align} \sum_{i=1}^\infty i^2\mathbb{P}(i\leq X_n<i+1) &= \mathbb{P}(X_n \geq 1) - \mathbb{P}(X_n\geq 2) + 2^2 \mathbb{P}(X_n \geq 2) - \mathbb{P}(X_n\geq 3) + \dots \\ &= \sum_{i=1}^\infty (2i-1)\mathbb{P}(X_n \geq i) \\ \mathbb{P}(X_n \geq n) & = \sum_{i=n}^\infty \mathbb{P}(i\leq X_n<i+1) \end{align}

However, none of these lead me to an answer. If it requires the Kolmogorov Zero-One Law, please explain me a little. I am confused about the definition of "tail field".

  • $\begingroup$ Is there are typo in the statement of the problem? Should it say, $\cdots \le C < \infty,$ and not $\cdots \le C \le \infty$? $\endgroup$ – Dfrtbx Jan 5 '18 at 17:45
  • $\begingroup$ @Dfrtbx It must be $C<\infty$ such that the tail sum $\sum_{i=k}^{k+m} i^2 P(X_n \geq i) \to 0$ for large enough $k$ as $m\to \infty$, which is what we prefer for the proof. Yet, only state close to $0$ is not small enough. It has to go to $0$ fast, which requires some algebra I'm missing. $\endgroup$ – Jango Jan 5 '18 at 18:06

The event $X_n \ge n$ i.o. is indeed a tail event. You can tell because if you change any finite number of the $X_n,$ it won't change the truth value of $X_n \ge n$ i.o. By the Kolmogorov 0-1 law, $\mathbb{P}(X_n \ge n \text{ i.o})$ is either 0 or 1. Therefore, it suffices to show that it is not equal to 1.

So assume that $\mathbb{P}(X_n \ge n \text{ i.o}) = 1.$ For infinitely many $n,$ we have that $$\sum_{k=1}^\infty k^2\mathbb{P}(k \le X_n < k+1) = \sum_{k=n}^\infty k^2\mathbb{P}(k \le X_n < k+1)\ge n^2,$$ contrary to hypothesis. $\square$

For more information on tail events and tail fields, see this answer.


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.