Does converge in distribution imply limit being finite almost surely?

262 Views Asked by At

If I have a sequence of random variable $x_n$ converge in distribution to a standard normal random variable, i.e.,

$$x_n\overset{d}{\to}N(0,1)$$

Does the following hold? Why or why not?

$$p(\lim_{n\to\infty}x_n<\infty)=1$$


Edited on 02/22/2018

More specifically, think about $x_n=\dfrac{1}{\sqrt{n}}\sum_{i=1}^{n}y_{I}$ where $y_i$'s are i.i.d. rv with zero mean and variance 1. By CLT you would have

$$x_n\overset{d}{\to}N(0,1)$$

The reason that I care about $p(\lim_{n\to\infty}x_n<\infty)=1$ is because that I am trying to find the limit of the following integral

$$\int f_n(t,x_n) dt$$

and I can show that $\lim_{n\to\infty}f_n(t,x_n)=f_{0}(t, x_n)$ and $f_{0}(t,x_n)$ is integrable no matter what $x_n$ is. I want to apply DCT and I have been able to show that for sufficiently large $n$,

$$f_n(t,x_n)<g(t)\exp\{x_n\},$$

where $g(t)$ is integrable. Hence it seems to me that as long as $\int g(t)\exp\{x_n\} dt<\infty$ holds almost surely I can apply DCT.

1

There are 1 best solutions below

0
On BEST ANSWER

The Law of Iterated Logarithm: https://en.wikipedia.org/wiki/Law_of_the_iterated_logarithm show that the opposite is true: $P\{\limsup x_n =\infty\} =1$