Show that $\liminf \frac{\vert S_n\vert}{\sqrt{n}} = 0$.

112 Views Asked by At

Let $X_1, X_2, \dots$ be i.i.d random variables with $\mathbb{E}(X_i) = 0$ and $0 < \operatorname{Var}(X_i) < \infty$, then by the Central Limit Theorem and Kolmogorov’s 0-1 Law we can conclude that $$\limsup_{n\rightarrow\infty} \frac{S_n}{\sqrt{n}} = \infty, \ a.s.$$

Now my question is that can we establish that $$\liminf_{n\rightarrow\infty} \frac{\vert S_n\vert}{\sqrt{n}} = 0, \ a.s.$$

It seems that the approach to the problem above fails here. I also tried to use the law of iterated logarithm to solve it, but only ended up with failure. Any help will be greatly appreciated.

1

There are 1 best solutions below

0
On

Yes, intuitivelly because the modulus of Gaussian $\mathcal N(0,1)$ is concentrated on the whole $[0,\infty)$.

Precisely; assume contrary that $\liminf \frac{|S_n|}{\sqrt{n}} \neq 0$. By Kołmogorov $0-1$ Law we can ensure that for some $\varepsilon > 0$, $\liminf \frac{|S_n|}{\sqrt{n}} > \varepsilon$ almost surely. Then, by Fatou lemma, $$ 1=\mathbb P(\liminf \frac{|S_n|}{\sqrt{n}} > \varepsilon) = \mathbb E[ 1_{\{\liminf \frac{|S_n|}{\sqrt{n}}> \varepsilon\}}] \le \mathbb E[\liminf_n 1_{\{\frac{|S_n|}{\sqrt{n}} > \varepsilon\}}] $$ $$ \le \liminf_n \mathbb E[1_{\{\frac{|S_n|}{\sqrt{n}} > \varepsilon\}}] = \mathbb P(|\mathcal N(0,Var(X_1))| > \varepsilon) = 1 - 2\Phi(\frac{\varepsilon}{\sqrt{Var(X_1)}}) < 1.$$