Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

I am trying to show giving a sequence $(X_n) \subseteq L^2$ of independent random variables with zero mean and $\sum_{n\in N}E[X_n^2]< \infty$ , $\sum_{n\in N}X_n$ converges a.s.

For this I wanna get $\sum_{n\in N} \mathbb{P}(|\sum_{k=1}^{n}X_k|>\epsilon)<\infty$. As if I can show it , then by Borel-Cantelli lemma, $\sum_{n\in N}X_n$ converges to zero almost surely.

But I dont know how to get $\sum_{n\in N} \mathbb{P}(|\sum_{k=1}^{n}X_k|>\epsilon)<\infty$,

I tried this:

By Chebychev's inequality, $\mathbb{P}(|\sum_{k=1}^{n}X_k|>\epsilon)\leq \mathbb{E}[X_n^2]/\epsilon^2 $, follows that $\sum_{n\in N}\mathbb{P}(|\sum_{k=1}^{n}X_k|>\epsilon)\leq\sum_{n\in N}\mathbb{E}[X_n^2]/\epsilon^2$. However, letting $\epsilon\longrightarrow 0$ doesn't make the right side bounded. Seems doesnt work...

I am wondering if anyone can show the correct way to do it. Thanks in advance.

share|cite|improve this question
I don't understand: if you can show what you said after the "I can get", then almost surely, $\sum_nX_n=0$. But it doesn't need to be the case. – Davide Giraudo Aug 13 '12 at 15:32
I need some help to prove that $\sum_{n\in N} \mathbb{P}(|\sum_{k=1}^{n}X_k|>\epsilon)<\infty$ – lindamac Aug 13 '12 at 16:31
@lindamac This inequality is false. That's why you can't prove it. – Byron Schmuland Aug 13 '12 at 16:40

We can apply Kolmogorov three series theorem. We have to check that

  1. $\sum_{j=1}^{+\infty}P(|X_j|>1)$ is convergent. It's the case, as $P(|X_j|>1)\leq EX_j^2$.
  2. $\sum_{j=1}^{+\infty}E[X_j\chi_{\{X_j\leq 1\}}]$ is convergent. We have, as $EX_j=0$, $$E[X_j\chi_{\{X_j\leq 1\}}]=-E[X_j\chi_{\{|X_j|>1\}}]\leq \sqrt{EX_j^2}\sqrt{P(|X_j|\geq 1)}\leq EX_j^2.$$

  3. $\sum_{j=1}^{+\infty}\operatorname{Var}(X_j\chi_{\{|X_j|\leq 1\}})$ is convergent. But since the expectation of $X_j $ is $0$, $$\operatorname{Var}(X_j\chi_{\{|X_j|\leq 1\}})\leq \operatorname{Var}(X_j)=EX_j^2,$$ and we can conclude.

share|cite|improve this answer
Sorry I've edited the question. Is there any way to use law of large numbers and Borel-Cantelli to solve the question , thought it might be longer than use three seriez theorem? – lindamac Aug 13 '12 at 16:25

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.