4
$\begingroup$

Let $Y_1$, $Y_2$, ... be a sequence of random variables such that $P(Y_n=\frac{1}{n})=1-\frac{1}{n^2}$ and $P(Y_n=n)=\frac{1}{n^2}$. Does $Y_n$ converge in probability?

I am stuck because I don't know how to deal with RV with two outcomes. And I am confused with the different probability in which $Y_n=1/n$ and $Y_n=n$. I know how to prove convergence with a single $Y_n$ to Y. But really stuck in this one.

so $$ Y_1=\begin{cases} 1 &\mbox{p } = 0 \\ 1 &\mbox{p } = 1 \end{cases} $$ $$ Y_2=\begin{cases} 2 &\mbox{p } = \frac{1}{4} \\ \frac{1}{2} &\mbox{p } = \frac{3}{4} \end{cases} $$

and then I don't know how to prove whether $Y_n$ converges or not in probability.

any help will be great. thanks

$\endgroup$
3
  • $\begingroup$ Use the definition of convergence in probability. $\endgroup$ Feb 2, 2020 at 6:44
  • $\begingroup$ When $n=1$, your equations yield $P(Y_1=1)=1$ and $P(Y_1=1)=0$, which is a contradiction. What does this mean? $\endgroup$
    – gunes
    Feb 2, 2020 at 15:52
  • $\begingroup$ Yes, this is exactly where I am confused with. But just like the two outcomes of each $Y_n$, the sum of the probability is 1. $\endgroup$ Feb 2, 2020 at 17:47

1 Answer 1

3
$\begingroup$

If we redesign your sequence as $Y_1=1$, and everything else remains the same, the contradiction disappears. Then, if $Y_n$ converges, it should converge to $0$. Here, we'll use the definition of convergence in probability we need to show the following: $$\lim_{n\rightarrow\infty}P(|Y_n-0|>\epsilon)=\lim_{n\rightarrow\infty} P(Y_n>\epsilon)=0$$

$Y_n$ is either $n$ or $1/n$. Now, let's fix some $n$, and calculate the above probability:

$$\begin{align}P(Y_n>\epsilon)&=P(Y_n>\epsilon|Y_n=n)P(Y_n=n)+P(Y_n>\epsilon|Y_n=1/n)P(Y_n=1/n)\\&=P(n>\epsilon)\frac{1}{n^2}+P(1/n>\epsilon)\left(1-\frac{1}{n^2}\right)\\&=\mathbb{I}(n>\epsilon){1\over n^2}+\mathbb{I}(1/n>\epsilon)\left(1-\frac{1}{n^2}\right)\end{align}$$

Now, we take the limit: $$\begin{align}\lim_{n\rightarrow\infty} P(Y_n>\epsilon)&=\underbrace{\lim_{n\rightarrow\infty}\mathbb{I}(n>\epsilon)}_{1}\underbrace{\lim_{n\rightarrow\infty}\frac{1}{n^2}}_0+\underbrace{\lim_{n\rightarrow\infty}\mathbb{I}(1/n>\epsilon)}_0\underbrace{\lim_{n\rightarrow\infty}\left(1-\frac{1}{n^2}\right)}_1\\&=0\end{align}$$

which is what we needed to prove.

$\endgroup$
2
  • $\begingroup$ Thanks a lot! This looks really great. I totally missed the conditional probability part. $\endgroup$ Feb 2, 2020 at 18:52
  • 2
    $\begingroup$ yes. sorry, I'm kinda new to this. $\endgroup$ Feb 2, 2020 at 19:46

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.