Suppose we have the following random variable:

$X_n = n$ with probability $\frac{1}{n}$ and $0$ with probability $1-\frac{1}{n}$. We we can define this variable on the probability space $([0,1], \mathcal{B}[0,1], \lambda)$ where $\lambda$ is the Lebesgue measure. Also, we have independence for all $n\geq1$.

It is said that this $X_n$ converges to $0$ almost surely (several sources). However, once we check this by the Borel - Cantelli Lemma 2, we get that $\sum_{n=1}^\infty P(X_n=n)=\infty$. Given that the events are independent, we know that $X_n=n$ infinitely often.

What could be the reason I am receiving this contradiction?

  • $\begingroup$ The source is correct, and never asserts that $(X_n)$ is independent (a hypothesis which you seem to have invented by yourself and which does not hold in the setting of the notes you are referring to). $\endgroup$ – Did Dec 17 '18 at 10:41
  • $\begingroup$ @Did I used the independence, because they do claim it in the same note later. It may have been a typo in the note. However, could you elaborate why the independence does not hold in this setting? $\endgroup$ – Ovi Dec 17 '18 at 10:52
  • $\begingroup$ Because they explicitely build $(X_n)$ as $X_n=\mathbf 1_{(0,1/n)}$ on $[0,1]$ endowed with its Borel sigma-field and the Lebesgue measure, and these are not independent. "they do claim it in the same note later" Sure they do, but for completely different sequences $(X_n)$. No typo here. $\endgroup$ – Did Dec 17 '18 at 11:44
  • $\begingroup$ @Did Thank you, I completely got it now! $\endgroup$ – Ovi Dec 17 '18 at 11:54

This is an answer to the question as posted here. The random variables are not defined explicitly in this question. It is just mentioned that they are defined on $(0,1)$ with Lebesgue measure. Explicit definition was given in the comments after I posted this answer.

I don't know what those sources are but we can only conclude that $X_n \to 0$ in probability. It need not converge almost surely.

  • $\begingroup$ Thanks for the comment! Here is the link: nptel.ac.in/courses/108106083/lecture28_Convergence.pdf). Page 3, Example. I also think that, clearly, we can only talk about convergence in probability. $\endgroup$ – Ovi Dec 17 '18 at 10:39
  • $\begingroup$ See comment on main. $\endgroup$ – Did Dec 17 '18 at 10:41

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.