2
$\begingroup$

Problem

Prove that $X_n \sim N(0,\frac{1}{n})$ converges in probability to 0.

Discussion

There a problems similar to this question, but I want to ask specifically about the convergence of an integral that arises in the solution.

Attempt

Let $\epsilon>0$. To show $X_n \overset{p}{\to} 0$, we must show $P(|X_n| > \epsilon) \to 0$.

$$ \begin{align} P(|X_n|>\epsilon) &= P(X_n>\epsilon) + P(X_n < -\epsilon)\\ &= P(Z>\epsilon\sqrt{n}) + P(X_n<-\epsilon\sqrt{n}) \end{align} $$ where $Z$ is the standard normal distribution. Thus we want to show $$P(|X_n|>\epsilon)=\int_{\epsilon\sqrt{n}}^\infty\frac{1}{\sqrt{2\pi}}e^{-\frac{1}{2}x^2}\text{dx} + \int_{-\infty}^{-\epsilon\sqrt{n}}\frac{1}{\sqrt{2\pi}}e^{-\frac{1}{2}x^2}\text{dx}\to 0$$

I understanding intuitively that these integrals should converge to zero, but how can I show rigorously that they go to zero?

$\endgroup$
1
$\begingroup$

You can use the tail bound for a standard normal random variable: $$ \int_x^\infty \exp(-y^2/2)\,dy \le x^{-1}\exp(-x^2/2), $$ which can be proved changing variables $y = x + z$ and using $\exp(-z^2/2) \le 1$.

$\endgroup$
  • $\begingroup$ I am selecting this as my answer because it directly shows how to work with the integral, which is what I was specifically asking about. The other answers also offer great solutions. $\endgroup$ – StrugglingStudent42 Dec 4 '20 at 17:25
2
$\begingroup$

Your integrals contain spurious instances of $\sqrt n$. You are evaluating $P(|Z|>\epsilon\sqrt n)$ where $Z$ is standard normal, so the integrand should not involve $\sqrt n$. Anyhow, you can prove $P(|Z|>\epsilon\sqrt n)\to0$ rigorously using Markov's inequality.

$\endgroup$
2
$\begingroup$

$$E[X_n]=0$$

$$\lim_{n \to \infty}\mathrm{Var}[X_n]=0$$

This is a necessary and sufficient condition for

$$X_n\xrightarrow{L^2}0$$

Convergence in $L^2$ implies convergence in probability

$\endgroup$

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.