0
$\begingroup$

Does $X_n$ with $\mathbb{P}[X_n = \frac{1}{n}] = 1 - \frac{1}{n^2}$ and $\mathbb{P}[X_n = n] = \frac{1}{n^2}$ converges in probability? In $L^2$?

Aplying limit definition I see that it is convergent. But I don't understand what I need to do in case of $L^2$ because I think that $L^2$ is for vectors and I don't see vectors here.

EDIT:

We need to prove that there exists random variable $X$ such that for any $\varepsilon > 0$ and any $\delta > 0$ there exists number $N$ (which may depend on one or both) such that for all $n \ge N, \Pr\left(|X_n - X| > \varepsilon\right) \lt \delta$ (definition of limit).

Let $X = 0$, then if we choose $N$ such that $\frac{1}{N} \lt \varepsilon$ and $\frac{1}{N^2} \lt \delta$ we will have that either $X_n \lt \varepsilon,$ or probability that this is false is less than $\delta.$ Thus $X_n$ is convergent in probability.

$\endgroup$
1
1
$\begingroup$

Actually, $X_n \to 0$ almost surely. This requires Borel Cantelli Lemma.

$\sum P(X_n=n)=\sum \frac 1 {n^{2}} <\infty$. By Borel-Cantelli Lemma it follows that $X_n=n$ i.o with probability $0$. Hence, $X_n=\frac 1 n$ for all $n$ sufficiently large, with proability $1$. It follows that $X_n \to 0$ a.s.

$EX_n^{2}=n^{2}(\frac 1{n^{2}})+\frac 1{n^{2}}(1-\frac 1 {n^{2}})\to 1$. If at all $(X_n)$ converges in $L^{2}$, it can only converge to $0$ because a.s. limit is $0$. It follows that $X_n$ does not converge in $L^{2}$.

$\endgroup$
3
  • $\begingroup$ The confusion seems to be more about what $L^2$ means, based on the OP's post; not so much about what to do from there. $\endgroup$
    – Clement C.
    Mar 28 at 23:27
  • 1
    $\begingroup$ Thank you, +1. I updated the answer with my simple proof of convergence. I'm not sure why do you need this lemma. $\endgroup$
    – Yola
    Mar 29 at 1:17
  • $\begingroup$ @Yola I thought you were asking for a.s. convergence, which is also true. I have edites my answer accordingly. $\endgroup$ Mar 29 at 5:03
2
$\begingroup$

I think you are confusing $\ell_2$ norms for vectors and sequences: $$ \|x\|_2 = \sum_{n=1}^\infty |x_n|^2 $$ and $L^2$ norms for functions (or, here, random variables: it's the same): $$ \|X\|_2 = \sqrt{\mathbb{E}[|X|^2]} $$ There definitely are connections between the two, these are not really different concepts.


In your case, $$ \|X_n\|_2^2 = \mathbb{E}[|X_n|^2] = \frac{1}{n^2}\cdot \left(1-\frac{1}{n^2}\right) + n^2\cdot \frac{1}{n^2} > n^2\cdot \frac{1}{n^2} = 1 $$ so $\|X_n\|_2^2 \not\to 0$. $X_n$ does not converge to $0$ in $L^2$.

$\endgroup$

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.