5
$\begingroup$

Let $X_i$ be i.i.d. $f_X(x)=\dfrac{1}{|x|^3}1_{\{|x|>1\}}$. Show that $\dfrac{\sum_{i=1}^nX_i}{\sqrt{n\log n}}\to N(0,1)$.

I realized that $Var(X_i)=\infty$ and $E(X_i)=0$. I cannot apply Lindeberg or Lyapounov CLT in any way, so I tried to use characteristic functions.

I observe that $$\lim_{n\to\infty}\varphi_{S_n}\left(\dfrac{t}{\sqrt{n\log n}}\right)=\exp\left[\lim_{n\to\infty}-\left (\int_1^\infty\dfrac{4n}{x^3}\sin^2 \left(\dfrac{tx}{2\sqrt{n\log n}}\right)dx\right)\right]$$

I now want to show that this limit is equal to $\exp(-t^2/2)$. How can I do that? I don't think I can apply Dominated Convergence here.

So essentially the task is to prove that the integral converges to $t^2/2$ as $n\to\infty$.

$\endgroup$
5
$\begingroup$

To use characteristic functions is definitely the way to go. We have, through integration by parts: $$ \int \frac{\sin^2(t x)}{x^3}\,dx = t^2\text{Ci}(2tx)-\frac{t}{2x}\sin(2tx)-\frac{1}{2x^2}\sin^2(tx)\tag{1}$$ hence for large $n$s: $$\int_{1}^{+\infty}\frac{4n}{x^3}\,\sin^2\left(\frac{tx}{2\sqrt{n\log n}}\right)\,dx \approx \frac{1}{2\log(n)}\left[3t^2-2t^2\text{Ci}\left(\frac{t}{\sqrt{n\log n}}\right)\right]\tag{2}$$ and the problem boils down to computing: $$ \lim_{n\to +\infty}\frac{3-2\,\text{Ci}\left(\frac{t}{\sqrt{n\log n}}\right)}{\log(n)}\tag{3} $$ that by De l'Hopital theorem equals: $$\lim_{n\to +\infty}\left(1+\frac{1}{\log n}\right)\cos\left(\frac{t}{\sqrt{n\log n}}\right) = 1.\tag{4}$$

$\endgroup$

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.