# expected value calculation for squared normal distribution

I need help with the following problem. Suppose $Z=N(0,s)$ i.e. normally distributed random variable with standard deviation $\sqrt{s}$. I need to calculate $E[Z^2]$. My attempt is to do something like \begin{align} E[Z^2]=&\int_0^{+\infty} y \cdot Pr(Z^2=y)dy\\ =& \int_0^{+\infty}y\frac{1}{\sqrt{2\pi s}}e^{-\frac y{2s}}dy\\ =&\frac{1}{\sqrt{2\pi s}}\int_0^{\infty}ye^{-\frac y{2s}}dy. \end{align}

By using integration by parts we get

$$\int_0^{\infty}ye^{-\frac y{2s}}dy=\int_0^{+\infty}2se^{-\frac y{2s}}dy=4s^2.$$

Hence $E[Z^2]=\frac{2s\sqrt{2s}}{\sqrt{\pi}},$ which does not coincide with the answer in the text. Can someone point the mistake?

-
Note that $\Pr(Z^2=y)=0$, identically. In fact you might want to debunk thoroughly the misconceptions which led you to write the identity where this quantity appears. – Did Mar 23 '13 at 7:55
Any luck with my suggestion above? – Did Mar 29 '13 at 7:58

This is old, but I feel like an easy derivation is in order.

The variance of any random variable $X$ can be written as $$V[X] = E[X^2] - (E[X])^2$$

Solving for the needed quantity gives $$E[X^2] = V[X] + (E[X])^2$$

But for our case, $E[X] = 0$, so the answer of $\sigma^2$ is immediate.

-

The answer is $s = \sigma^2$. The integral you want to evaluate is

$$E[Z^2] = \frac{1}{\sqrt{2 \pi} \sigma} \int_{-\infty}^{\infty} dz \: z^2 \exp{(-\frac{z^2}{2 \sigma^2})}$$

-
The answer should be $s$ because $s=\sigma^2$. I am curios what is wrong with the method I wrote. – John Peter Dec 23 '12 at 10:54
Your method is wrong because you misunderstand the definition of expected value. The expected value of a random variable $Z$ having probability distribution $f(z)$ is $\int_{-\infty}^{\infty} dz \: z f(z)$. For the expected value of a function $g(Z)$, the result is $\int_{-\infty}^{\infty} dz \: g(z)f(z)$. – Ron Gordon Dec 23 '12 at 11:03