2
$\begingroup$

I got a bit confused during the end of this proof so I am asking for a check. Take $$Y(n) = \begin{cases} 1 &\mbox{with probability} \ 1 -p_n \\ n & \mbox{with probability} \ p_n \end{cases} $$

Assume $p_n \rightarrow 0$ prove that $Y_n$ converges in probability to 1.

My proof:

We want to show that $\lim_{n \rightarrow \infty} P(|Y_n -1| < \epsilon) = 1$, notice that $P(|Y_n -1| < \epsilon) = P(Y_n < \epsilon+1 \ \mbox{or} \ -\epsilon+1 < Y_n \ )$.

In the case $P(-\epsilon+1 < Y_n \ )$ we have that $P( Y_n > -\epsilon+1 \ ) = 1-p_n +p_n= 1 \ \forall{n}$. So here we are done.

In the case $P(Y_n < \epsilon+1)$ we see that $\forall\epsilon \ge 1$ taking $N_1 = \left \lfloor{\epsilon +1}\right \rfloor $ will ensure that $\forall n > N_1 \ P(Y_n < \epsilon+1)= 1-p_n +p_n= 1 $. If $\epsilon$ is less than one then $P(Y_n < \epsilon+1) = 1-p_n$ but we know that there exists an $N_2$ s.t. $\forall n > N_2 \implies p_n < \epsilon_2$ so choosing $N = \max \{ N_1, N_2 \}$ will make $$|P(Y_n < \epsilon+1) -1| < \epsilon_2$$ $\forall n > N.$

Other ways of solving this are welcome, I have a feeling I made it longer than it needed to be.

$\endgroup$

2 Answers 2

2
$\begingroup$

Most of the intermediate work can be skipped. You don't need to consider all possible $\epsilon$--you are always permitted to focus on sufficiently small $\epsilon$. Similarly, you may ignore a finite number of $n$. So assume $0 \lt \epsilon \lt 1$ and restrict the sequence to $n\ge 2$, whence

$$\lim_{n\to\infty}\Pr(|Y_n - 1| \lt \epsilon) =\lim_{n\to\infty}\Pr(Y_n = 1) = \lim_{n\to\infty}1 - p_n = 1,$$

QED.

$\endgroup$
1
$\begingroup$

The idea is correct. I think the formula $$\forall n > N_1 \ P(Y_n < \epsilon+1)= 1-p_n +p_n= 1$$ should read $$\forall n > N_1 \ P(Y_n < \epsilon+1)= 1-p_n,$$ because for such $n$'s, $\{Y_n < \epsilon+1\}=\{Y_n=1\}$.

An "other" way could be the following: define $Z_n:=|Y_n-1|$; then $\mathbb P(Z_n=0)=1-p_n$ and $\mathbb P(Z_n=n-1)=p_n$. Fix $\varepsilon\in (0,1)$. Then for $n \geqslant 3$, we have $$\{|Z_n|> \varepsilon \} \subset\{Z_n=n-1\},$$ hence $$\mathbb P\{|Z_n|> \varepsilon \} \leqslant \mathbb P\{Z_n=n-1\}=p_n.$$

$\endgroup$

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.