3
$\begingroup$

I'm stuck with this problem: Calculate $\mathbb{E}[N(X)]$, where N(·) is the cdf of the standard normal distribution, and X is a standard normal random variable.

Here's where I'm stuck: We know that:

$$\mathbb{E}[g(X)]=\int_{-\infty }^{\infty}g(x)f_{X}(x)dx$$

And we are given that:

$$ g(x) = N(x) = \frac{1}{\sqrt{2 \pi}} \int_{-\infty}^{x}e^{-\frac{u^2}{2}}du $$

$$ f_{X}(x)=\frac{1}{\sqrt{2 \pi}} e^{-\frac{x^2}{2}} $$

So, the solution should be to develop this:

$$\begin{split} \mathbb{E}[N(X)] &= \int_{-\infty }^{\infty} \left( \frac{1}{\sqrt{2 \pi}} \int_{-\infty}^{x}e^{-\frac{u^2}{2}}du \right) \left( \frac{1}{\sqrt{2 \pi}} e^{-\frac{x^2}{2}} \right) dx\\ &= \frac{1}{2 \pi} \int_{-\infty }^{\infty} \left( \int_{-\infty}^{x}e^{-\frac{u^2}{2}}du \right) e^{-\frac{x^2}{2}} dx\\ &= \frac{1}{2 \pi} \int_{-\infty }^{\infty} \int_{-\infty}^{x} e^{-\frac{u^2+x^2}{2}}dudx \end{split}$$

My questions are:

  1. Can we combine the exponents moving the term with x inside, the last step above?
  2. If yes, this looks like a good opportunity to change variables to polar. If that's the case, how to change the limits of the integrals from $x$ and $u$ to $r$ and $\theta$?

Any help would be extremely appreciated! Thanks!

$\endgroup$
3
  • 2
    $\begingroup$ Your last step is reasonable, but this is not the fastest approach. Instead note that when $0 < N(x) <1$ you have $\mathbb P(N(X) \le N(x))=\mathbb P(X\le x)= N(x)$ as you have a continuous distribution. When $\mathbb P(Y \le y) =y$ for $0 <y<1$ you have a uniform distribution on $[0,1]$. $\endgroup$
    – Henry
    Commented Mar 4, 2023 at 0:52
  • $\begingroup$ A relevant question for you to check: stats.stackexchange.com/q/161635/328953 . Another relevant one with the proof: math.stackexchange.com/q/868400/915791 $\endgroup$ Commented Mar 4, 2023 at 0:55
  • $\begingroup$ Alternatively, the function $g(x,u)=\exp(-(u^2+x^2)/2)$ statistics $g(x,u)=g(u,x),$ so the integral over your region, where $u<x,$ is exactly half of the integral across the whole plane, which is $1.$ So you get $1/2.$ The would work all the time, for continuous random variable, where you get $g(x,u)=f_X(u)f_X(x).$ $\endgroup$ Commented Mar 4, 2023 at 1:03

3 Answers 3

1
$\begingroup$

@Golden_Ratio's answer is the most general as it holds for an arbitrary random variable. However, the problem at hand can be easily determined by denoting $\Phi$ and $\phi$ as the standard normal cdf/pdf and then writing $$ \mathsf E\Phi(X)=\int_{-\infty}^\infty\Phi(x)\phi(x)\,\mathrm dx=\int_{-\infty}^\infty\Phi(x)\,\mathrm d\Phi(x)=\frac{\Phi^2(x)}{2}\bigg|_{-\infty}^\infty=\frac{1}{2}. $$

$\endgroup$
0
$\begingroup$

Let $X$ be a random variable having invertible CDF $F$. Then the random variable $F(X)$ is uniformly distributed on $(0,1)$ since

$$ P(F(X)\leq y)=P(X\leq F^{-1}(y)) =F(F^{-1}(y))=y.$$

So it immediately follows that $E[F(X)]=1/2.$

$\endgroup$
0
$\begingroup$

$A(x)=\int\limits_{-\infty}^xe^{-\frac{u^2}{2}}du$ and $dA=e^{-\frac{x^2}{2}}dx$ so integral becomes $\frac{1}{2\pi}\int\limits_0^{\sqrt{2\pi}}AdA=\frac{1}{2}$.

$\endgroup$
2
  • $\begingroup$ Thanks, this is also a great approach! Question: why does the superior limit becomes $\sqrt{2 \pi}$? $\endgroup$
    – user1146971
    Commented Mar 4, 2023 at 18:07
  • $\begingroup$ When $x=\infty$, $A(\infty)=\sqrt{2\pi}$. $\endgroup$ Commented Mar 4, 2023 at 23:15

You must log in to answer this question.