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

On wikipedia i have find this statement: is scale invariant, and the only continuous distribution that fits this (scale invariance) is one whose logarithm is uniformly distributed.

how can be proven?

share|cite|improve this question
up vote 2 down vote accepted

This is the argument as it's sketched by Julian Havil in Gamma: Exploring Euler's Constant; some fiddly details (e.g., involving domains of definition) are ignored.

Let $X$ be a scale-invariant continuous random variable with probability density function $f(x)$ and cumulative density function $F(x)$. Fix a lower bound $u$ on $X$. Scale invariance means that for any $a>0$, $$P(u<X<x)=P\left(u<\frac1aX<x\right)=P(au<X<ax)\;,$$ and hence $F(ax)-F(au)=F(x)-F(u)$. Differentiating with respect to $x$, we see that $af(ax)=f(x)$, so $$f(ax)=\frac1af(x)\;.\tag{1}$$

Now fix a base $b$ for logarithms and let $Y=\log_b X$, with pdf $g(y)$ and cdf $G(y)$. Then $$G(y)=P(Y\le y)=P(\log_bX\le y)=P(X\le b^y)=F(b^y)=F(x)\;,$$ and hence

$$\begin{align*} g(y)&=\frac{d}{dy}G(y)=\frac{d}{dy}F(x)\\\\ &=\frac{d}{dx}F(x)\frac{dx}{dy}=f(x)\frac{dx}{dy}\\\\ &=xf(x)\ln b\;.\tag{2} \end{align*}$$

Substitution of $ax$ for $x$ and using $(1)$ and $(2)$ yields

$$\begin{align*}g(\log_b ax)&=axf(ax)\ln b\\ &=ax\cdot\frac1af(x)\ln b\\\\ &=xf(x)\ln b\\\\ &=g(\log_b x)\;, \end{align*}$$

which can be rewritten as $g(\log_b x+\log_b a)=g(\log_b x)$. But $a$ was an arbitrary positive scaling factor, so $g(y+c)=g(y)$ for all $c\in\Bbb R$, and therefore $g$ must be a constant function.

This shows that if $X$ is continuous and scale-invariant, then $\log_bX$ is uniformly distributed.

share|cite|improve this answer
a comment: in a lot of books i have seen scale invariance defined as $f(ax)=a^k f(x)$. In this case $\log(x)$ does not distribute uniformly i suppose. – emanuele May 6 '12 at 7:06
@emanuele: I'd have to work through the algebra, but it doesn't look like it. – Brian M. Scott May 6 '12 at 7:17

If you start by assuming that $X\sim\operatorname{Unif}[0,1]$ and then examine the leading digit of $Y=b^X$ base $b$, I think you will be able to derive the law. Let $B$ be the leading digit base $b$ (B for Benford). Then $B = \lfloor Y\rfloor$ and $$P\left(B=k\right)=\log_b(k+1)-\log_b k=\log_b\left(1+\frac1k\right)$$ follows from observing that $$k\le b^X\lt k+1\qquad\iff\qquad\log_bk\le X\lt \log_b(k+1)$$ and $X$ is uniform.

share|cite|improve this answer

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.