Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

I'm studying statistics and there's one part in my book I can't understand. I tried to make as good translation as I can of the problematic goes:

Chi squared $\chi^2$ distribution

Let $Z_{1}, ..., Z_{v}$ be $v$ independent $N(0,1)-$ distributed random variables and

$\chi_{v}^2 = Z_{1}^2 + ... + Z_{v}^2$.

If $v = 1$ the cumulative distribution function $F_{\chi_{1}^2}(x) = F_{1}(x)$ is the following:

$F_{1}(x) = P(Z_{1}^2 \leq x) = P(-\sqrt{x} \leq Z_{1} \leq \sqrt{x}) = 2\Phi(\sqrt{x}) - 1, $ where $x \geq 0$ and $\Phi(x)$ is the CDF of the normal distribution . By taking the derivative we get the probability density function to be

$f_{1}(x) = x^{-\frac{1}{2}}\phi(\sqrt{x}) = C_{1}x^{-\frac{1}{2}}e^{-\frac{1}{2}x}$, where $x \geq 0$ and $C_{1} = \frac{1}{\sqrt{2\pi}}$ is the normalizing constant.

Let's take a closer look of the sum $Z_{1}^2 + Z_{2}^2$ when $v = 2$. Both $Z_{1}^2$ and $Z_{2}^2$ are normally distributed. By calculating the CDF $P(Z_{1}^2 + Z_{2}^2 \leq x)$ as the integral of the product of density functions of $Z_{1}^2$ and $Z_{2}^2$ and then by taking the derivative we get

(1) $f_{2}(x) = \displaystyle\int_{0}^{x} f_{1}(x-u)f_{1}(u)du = C_{1}^2\displaystyle\int_{0}^{x}(x-u)^{-\frac{1}{2}}u^{-\frac{1}{2}}e^{-\frac{1}{2}x}du$

By making the substitution $u = xt$, we get easily

(2) $f_{2}(x) = C_{2}e^{-\frac{1}{2}x}$, where $C_{2}$ is again the normalizing constant.

Now my question is: "How did the writer of this book get the (1)-part of the $f_{2}(x)$-function. I didn't understand that part. Can someone give me more detailed steps or something? Thank you for any help

P.S. If any part is unclear or something I can provide more information on the part

share|cite|improve this question
up vote 1 down vote accepted

You can write the probability in term of expectation, with indicator functions:

$$P(Z_1^2 + Z_2^2 \leq x) = E[ 1_{Z_1^2 + Z_2^2 \leq x}]$$

$$ = \int_0^{+\infty} \int_0^{+\infty} 1_{u+v\leq x} f_1(u) f_1(v) du dv$$

$$ = \int_0^{+\infty} \int_0^{+\infty} 1_{u \leq x-v} f_1(u) du f_1(v) dv$$

$$ = \int_0^{+\infty} F_1(x-v) 1_{v \leq x} f_1(v) dv$$

$$ = \int_0^{x} F_1(x-v) f_1(v) dv$$

I added the indicator function $1_{v \leq x}$ in the 3rd line because $F_1(x-v)$ is only defined where $x \geq v$.

Also, you know that $f_2(x) = \frac{d}{dx} P(Z_1^2 + Z_2^2 \leq x)$


$$f_2(x) = \int_0^{x} f_1(x-v) f_1(v) dv$$

$$f_2(x) = C_1^2 \int_0^{x} \sqrt{u(x-u)} dv e^{-\frac{1}{2}x}$$

$$f_2(x) = C_1^2 \int_0^{1} \sqrt{t(1-t)} dt e^{-\frac{1}{2}x}$$

and we get the same result as the author.

share|cite|improve this answer
Thank you for your effort! =) – jjepsuomi Feb 17 '13 at 20:07

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

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