5
$\begingroup$

Let $Y$ be a continuous random variable with density function $f_Y(y)=e^{-y}, y > 0$. Consider the sequence $\{X_n\}$, given by $X_n=e^nI_{\{Y>n\}}, n =1,2,\cdots$

How can I show that $\{X_n\} \to 0$ in probability?

My work:

So, I want to show that $P(|X_n-0|>\epsilon) \to 0$ as $n \to \infty$.

$P(|X_n| > \epsilon)=P(X_n > \epsilon)=1-P(X_n \le \epsilon)=1-\int_0^\epsilon\Sigma_1^Y(e^n)dx$

However, I do not think my last equality makes much sense. How would you proceed?

| cite | improve this question | |
$\endgroup$
  • $\begingroup$ What's the $I$ here? $\endgroup$ – jkm Nov 23 '19 at 20:26
  • $\begingroup$ @jkm the indicator function $\endgroup$ – Ron Snow Nov 23 '19 at 20:42
  • $\begingroup$ Draw graphs of some of the $X_n.$ The rest will be obvious. $\endgroup$ – whuber♦ Nov 23 '19 at 22:53
4
$\begingroup$

Consider the following inequality for $\varepsilon \leq 1$

$$P( |X_n| > \varepsilon ) = P(e^n I_{\{Y>n\}}>\varepsilon) \leq P( I_{\{Y>n\}} > \varepsilon) = P(Y > n) = e^{-n} \to 0.$$

Actually we do not need to know the distribution of $Y$. We can use the cumulative function properties $$ P( Y > n) = 1 - F_Y(n) \underset{n\to \infty}{\to} 1 -1 = 0$$

For $\varepsilon > 1$ we get $$ P( |X_n| > \varepsilon) \leq P(|X_n| > 1) \to 0$$

| cite | improve this answer | |
$\endgroup$
  • $\begingroup$ why the downvote? $\endgroup$ – Manuel Nov 23 '19 at 20:56
  • $\begingroup$ I'll equal it out for you! I like this idea, but how can you justify the inequality and the equality thereafter? $\endgroup$ – Ron Snow Nov 23 '19 at 21:05
  • 1
    $\begingroup$ If $n \geq 1$ then $ e^n I_{Y > n} > I_{Y > n}$. Also An indicator is $1$ or $0$ so the probability of being greater than $\varepsilon$ is equal to the probability ob being 1. $\endgroup$ – Manuel Nov 23 '19 at 21:08
  • 1
    $\begingroup$ (+1) for the inequality in the first line. But note that you need to also consider the case $\epsilon\geq 1$ for completeness. $\endgroup$ – gunes Nov 23 '19 at 21:20
  • 1
    $\begingroup$ The last line "$X_n$ are positive with probability one": that's not correct, isn't it? $\endgroup$ – StijnDeVuyst Nov 23 '19 at 21:32
1
$\begingroup$

First, $\text{Prob}[Y>n]=e^{-n}$ since $Y$ has an exponential distribution. Therefore, $X_n$ can have only two values, namely $e^n$ with probability $e^{-n}$ and $0$ with probability $1-e^{-n}$. Let's fix some positive value $\epsilon$ and call $n'$ the smallest nonnegative integer bigger than $\ln \epsilon$. Then, for any $n \geqslant n'$, $\text{Prob}[X_n>\epsilon]=e^{-n}$, which clearly goes to 0 for large $n$.

| cite | improve this answer | |
$\endgroup$
  • $\begingroup$ I have several questions for you. (1) There are two values for $X_n$ due to the indicator function, correct? (2) Why isn't $P(X_n > \epsilon) = e^{-n} + (1 - e^{-n})$? $\endgroup$ – Ron Snow Nov 23 '19 at 21:09
  • 1
    $\begingroup$ (1) correct (2) $\epsilon$ can be small, but never equal to 0. Therefore $\text{Prob}[X_n>\epsilon]=\text{Prob}[X_n=e^n]=e^{-n}$. $\endgroup$ – StijnDeVuyst Nov 23 '19 at 21:17
1
$\begingroup$

$X_n$ can have two values (i.e. either $0$ or $e^n$) with $p=P(X=e^n)=e^{-n}$. For $\epsilon\geq e^{n}$, $P(X_n>\epsilon)$ is always $0$. For $\epsilon<e^n$, $P(X_n>\epsilon)=p=e^{-n}$, and this goes to $0$ as $n$ goes to infinity.

| cite | improve this answer | |
$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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