6
$\begingroup$

I posted this on mathstackexchange several days ago, but it may be more appropriate to ask this here.

Let $r_n$ be some real sequence converging to zero. Let $X_n$ be a sequence of random variables. I know that $X_n=O_p(r_n)$, which means that $\forall\varepsilon>0,\exists M$ such that $$\Pr(|X_n|>r_nM)<\varepsilon,\quad \forall n$$

Does this imply that $\forall \varepsilon>0$, there exists some finite constant $C$ such that $$\Pr(|X_n|>\varepsilon)\leq Cr_n?$$

It is clear that since $r_n\to 0$, $X_n\to_p0$, so that $\forall\varepsilon>0$ $$\lim_{n\to\infty}\Pr(|X_n|>\varepsilon)=0,$$ but I don't see how to obtain what I need.

$\endgroup$

1 Answer 1

1
$\begingroup$

Consider $X_n$ such that $\mathbb P\left(X_n=n\right)=1/n$ and $\mathbb P\left(X_n=0\right)=1-1/n$, and $r_n:=1/n^2$. Since the sequence $X_n/r_n$ converges to $0$ in probability, we have in particular $X_n=O_p(r_n)$. However, for each positive $\varepsilon$, the quantity $\mathbb P\left(X_n\gt\varepsilon\right)$ is equal to $1/n$ if $n\gt \varepsilon$.

$\endgroup$

Your Answer

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

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