2
$\begingroup$

Let $(X_n)_n$ be a sequence of iid random variables with $P(X_1 = 1) = \frac{1}{2}$ and $P(X_1 = -1) = \frac{1}{2}$. Define $S_n := \sum_{i = 1}^n X_i$. Prove that $S_n$ does not converge almost surely by contradiction using the Central Limit Theorem.

The CLT states that: $\frac{S_n-nE(X_1)}{\sqrt{Var(X_1)n}} = \frac{S_n}{\sqrt{n}} \overset{\mathcal{L}}{\longrightarrow} \mathcal{N}(0,1)$.

I don't really have an idea how to deduce that $S_n$ is not almost surely convergent using the CLT.

$\endgroup$

1 Answer 1

Reset to default
3
$\begingroup$

Suppose $S_n\to S$ almost surely for some random variable $S$. Then $$ \frac{S_n}{\sqrt{n}}\to 0\quad a.s. $$

One the other hand, almost surely convergence implies convergence in distribution. What contradiction can you have?

$\endgroup$
3
  • $\begingroup$ Thank you for the answer but why is $\frac{S_n}{\sqrt{n}} \to 0$ almost surely? The contradiction would be as you mentioned with the convergence in distrubution. According to the CLT $\frac{S_n}{\sqrt{n}} \overset{\mathcal{L}}{\longrightarrow} \mathcal{N}(0,1)$ whereas a constant random variable $c$, especially 0, has the distribution function $\mathbb{1}_{[c,\infty)}$. Hence, the necessary contradiction. $\endgroup$
    – PeterMcCoy
    Dec 18, 2016 at 22:50
  • $\begingroup$ Think about the case for real sequence. If $a_n\to a$ for some real number $a$, then what can you say about $\lim_{n\to\infty}\dfrac{a_n}{\sqrt{n}}$? $\endgroup$
    – user9464
    Dec 18, 2016 at 22:53
  • $\begingroup$ Oh well, of course you are right. $\lim_{n \to \infty} a_n/ \sqrt{n} = 0$ and this also applies for the almost surely convergence since you assumed that $S_n \to S$ a.s. $\endgroup$
    – PeterMcCoy
    Dec 18, 2016 at 22:57

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.