Relation between Big O and convergence in limit of a sequence of probabilities

297 Views Asked by At

I'm confused on the procedure used to show the theorem 5.52 in van der Vaart "Asymptotic Statistics" p.75. Here the simplified idea. Consider the sequence of real-valued positive random variables $\{X_n\}$ and suppose we want to show that

(*) $X_n \in O_p(1)$ : $\forall K>0$ $\exists M>0$ s.t. $P(X_n>M)<K$ $\forall n \in \mathbb{N}$

Suppose we show that

(**) $\lim_{n \rightarrow \infty} P(X_n>M)=0$ for some $M>0$: $\forall \epsilon>0$ $\exists \bar{n}\in \mathbb{N}$ s.t. $P(X_n>M)<\epsilon$ $\forall n \geq \bar{n}\in \mathbb{N}$

Does (**) imply (*)?

I think it doesn't because

(**) implies that: $\forall \epsilon>0$, $\exists A_{\epsilon}:=\max\{X_1,...,X_{\bar{n}-1}, \epsilon\}$ s.t. $P(X_n>M)<A_{\epsilon}$ $\forall n \in \mathbb{N}$.

However, the procedure used in the proof of the theorem 5.52 in van der Vaart "Asymptotic Statistics" p.75 seems to be based on the idea that (**) implies (*).

Could you help me to clarify this?