2
$\begingroup$

I'm working on a problem in which I should prove "almost sure" convergence for a sequence of random variables. I'm using Borel-Cantelli lemma to prove it. Here is the question and my solution - I appreciate if you can give me feedback and whether you have any other solution to this problem.

Assumptions:

  • $X_n$ is a sequence of random variables
  • $Pr(|X_n| > \epsilon)\le c^n$ for sufficiently large n, where $ 0 < c < 1$ and for $\forall \epsilon > 0$

Goal: to show:

$$X_n \rightarrow^{a.s} 0 \text{ a.s. stands for almost sure convergence}$$


My solution:

As $n \rightarrow \infty$, we can write:

$$\sum_{j = 1}^{\infty} Pr(|X_n| > \epsilon) \le \frac{c}{1-c} < \infty$$

Then by the application of Borel-Cantelli lemma, we can conclude:

$Pr\{|X_n| > \epsilon, i.o\} = 0 \Rightarrow 1 - Pr\{|X_n| > \epsilon, i.o\} = Pr\{|X_n| \le \epsilon, i.o\} \Rightarrow Pr\{|X_n| \le \epsilon, i.o\} = 1 $ $\Rightarrow X_n \rightarrow^{a.s} 0$

What do you think about this solution?

$\endgroup$
0
$\begingroup$

Careful, the complement of the set $\{|X_n|\leqslant\varepsilon, \mbox{i.o.}\}$ is not $\{|X_n|>\varepsilon, \mbox{i.o.}\}$ (it is $\{\exists m\mid \forall n\geqslant m,|X_n|>\varepsilon\}$.

So the application of Borel-Cantelli's lemma yields a set $\Omega_{\varepsilon}\subset \Omega$ of measure $1$ for which if $\omega\in\Omega'$, then there is an integer $N=N(\omega)$ such that $|X_n|\leqslant \varepsilon$ whenever $n\geqslant N$.

Take $\Omega':=\bigcap_{q\in\Bbb Q,q>0}\Omega_q$: there is pointwise convergence on $\Omega'$ to $0$ and $\mathbb P(\Omega')=1$.

Since we just need the convergence of $\mathbb P(|X_n|>\varepsilon)$, we actually proved that

If for each $\varepsilon>0$, the series $\sum_{n=1}^{+\infty}\mathbb P(|X_n|>\varepsilon)$ is convergent, then $X_n\to 0$ almost surely.

$\endgroup$

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.