2
$\begingroup$

Let $(X_j)_{j\in \mathbb Z}$ be an strictly stationary sequence of random variables with: $$E[X_j]=0, \quad E[X_j^2] < \infty$$ I want to show that for each positive $\varepsilon$: $$ \sum_{n=1}^\infty \mathbb P\left(\max_{1\leq j\leq 2^n}\lvert X_j\rvert>2^{n/2}\varepsilon\right)<\infty. $$ Note that: $$ \mathbb P\left(\max_{1\leq j\leq 2^n}\lvert X_j\rvert>2^{n/2}\varepsilon\right)= \mathbb P\left(\bigcup_{j=1}^{2^n} \left[ |X_j| >2^{n/2}\varepsilon\right]\right)\leq \sum_{j=1}^{2^n} \mathbb P\left( |X_j| >2^{n/2}\varepsilon\right) $$ By stationarity, we have $$\sum_{j=1}^{2^n} \mathbb P\left( |X_j| >2^{n/2}\varepsilon\right)= 2^n \mathbb P\left( |X_1|>2^{n/2}\varepsilon\right)= 2^n \mathbb P\left( |X_1|^2>2^{n}\varepsilon^2\right)$$ My initial idea is to use a comparison test (I accept another solution). However, I am unable to finish.

Could you help me finish?

$\endgroup$
3
  • 1
    $\begingroup$ If $E[X_j]=0$, then most people would assume that $E[X_j]^2=0$ also, and so the second inequality would be unnecessary.. Or did you mean to write $E[X_j^2] < \infty$ (which makes a lot more sense) instead of $E[X_j]^2 < \infty $? $\endgroup$ Jun 17 at 19:10
  • 1
    $\begingroup$ Can you add the source of this result? $\endgroup$
    – Zhanxiong
    Jun 18 at 3:39
  • $\begingroup$ Please, see the comment of the answer of this question. math.stackexchange.com/questions/4719818/… I spent a few hours thinking about the suggestion the respondent gave me, but I couldn't. I'm afraid it's something very trivial that I'm missing. Since I don't like to take too much time from one respondent, I decided to post this question here. $\endgroup$ Jun 18 at 3:49

1 Answer 1

2
$\begingroup$

To finish the proof, use the equality $E[X_1^2] = \int_0^\infty P[X_1^2 > t]dt$, whence \begin{align*} & E[X_1^2] = \int_0^\infty P[X_1^2 > t]dt \\ \geq & \sum_{n = 1}^\infty \int_{2^{n - 1}\varepsilon^2}^{2^n\varepsilon^2}P[X_1^2 > t]dt \\ \geq & \sum_{n = 1}^\infty 2^{n - 1}\varepsilon^2P[X_1^2 > 2^n\varepsilon^2]. \end{align*} Linking the summand in the last expression above to the last probability in your post and using the condition $E[X_1^2] < \infty$ finishes the proof.

This answer applied the similar trick.

$\endgroup$
1
  • $\begingroup$ Dear, I had already tried to do this: $$2^n \mathbb P\left( |X_1|^2>2^{n}\varepsilon^2\right)= \frac{\mathbb E [|X_1|^2] }{\varepsilon^2}$$ However, the series $\sum_{n=1}^\infty \frac{\mathbb E [|X_1|^2] }{\varepsilon^2}$ Does not converge. $\endgroup$ Jun 17 at 20:57

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.