19
$\begingroup$

I have gaussian distributed numbers with mean 0 and variance 0.2.

And I want to transform this distribution to uniform distribution [-3 3].

How can I transform gaussian distribution numbers to uniform distribution?

$\endgroup$

2 Answers 2

12
$\begingroup$

\begin{align} X & \sim N(0,0.2) = N\left( 0, \frac 1 5 \right) \\[10pt] X\cdot\sqrt 5 &\sim N(0,1) \\[10pt] \Phi(X\cdot\sqrt 5) & \sim \operatorname{Uniform}[0,1] \tag 1 \\[10pt] -3 + 6 \Phi(X\cdot \sqrt 5) & \sim \operatorname{Uniform}[-3,3] \end{align}

A proof of line $(1)$ is as follows: Suppose $Z\sim N(0,1).$ Then for $0 \le x\le 1,$ $$ \Pr(\Phi(Z) \le x) = \Pr( Z \le \Phi^{-1}(x)) = \Phi(\Phi^{-1}(x)) = x. $$

$\endgroup$
1
11
$\begingroup$

Let $X\sim \mathcal{N} (\mu, \sigma^2)$ have a normal distribution with mean $\mu=0$ and variance $\sigma^2 = 0.2$, which cumulative distribution function (CDF) is denoted by $\Phi_X$. The variable $Y = 6\Phi_X (X) - 3$ has a uniform distribution over $[{-3},3]$. In facts, \begin{aligned} \mathbb{P}(Y\leq t) &= \mathbb{P}\left(\Phi_X(X)\leq \frac{t+3}{6}\right) \\ &= \mathbb{P}\left(X\leq \Phi_X^{-1}\left(\frac{t+3}{6}\right)\right) \\ &= \Phi_X\left(\Phi_X^{-1}\left(\frac{t+3}{6}\right)\right) \\ &= \frac{t+3}{6} \, , \end{aligned} if ${-3}\leq t \leq 3$, $\mathbb{P}(Y\leq t)=0$ if $t \leq {-3}$, and $\mathbb{P}(Y\leq t)=1$ if $t \geq 3$.

$\endgroup$
3
  • $\begingroup$ You missed something: How is the fact that $\sigma^2 =0.2$ taken into account in your answer? $\endgroup$ Commented Jul 3, 2017 at 4:58
  • $\begingroup$ oh: You intended $\Phi$ to be the c.d.f. of a normal distribution other than $N(0,1).$ Nonstandard notation, but that makes your answer correct. $\endgroup$ Commented Jul 3, 2017 at 4:59
  • $\begingroup$ Thanks for your solution. Can you answer a follow-up question? math.stackexchange.com/questions/3406481/… $\endgroup$
    – Albert
    Commented Oct 24, 2019 at 0:36

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .