Let's consider $Y_n$ the max of $n$ iid samples $X_i$ of the same distribution:

$Y_n = max(X_1, X_2, ..., X_n)$

Do we know some common distributions for $X$ such that $Y$ is uniformly distributed $U(a,b)$?

I guess we can always "construct a distribution" $X$ to enforce this condition for $Y$ but I was just wondering if a famous distribution satisfies this condition.

  • 1
    $\begingroup$ It is worth noting that the law of any such $X$ must be unique, so "constructing a distribution" will give you the (only) answer. $\endgroup$ – πr8 Sep 3 '20 at 7:10
  • $\begingroup$ @πr8 can you elaborate why such a law has to be unique? $\endgroup$ – Philippe Remy Sep 3 '20 at 15:34
  • $\begingroup$ math.stackexchange.com/questions/1397427/… $\endgroup$ – shayaa Sep 3 '20 at 18:11
  • $\begingroup$ if a maximum $Y$ is a scalar, how can it have a distribution? $\endgroup$ – develarist Sep 4 '20 at 1:00
  • $\begingroup$ @develarist it's not a scalar. It's a sequence defined by $n$ and $X$. $\endgroup$ – Philippe Remy Sep 4 '20 at 11:50

Let $F$ be the CDF of $X_i$. We know that the CDF of $Y$ is $$G(y) = P(Y\leq y)= P(\textrm{all } X_i\leq y)= \prod_i P(X_i\leq y) = F(y)^n$$

Now, it's no loss of generality to take $a=0$, $b=1$, since we can just shift and scale the distribution of $X$ to $[0,\,1]$ and then unshift and unscale the distribution of $Y$.

So what does $F$ have to be to get $G(y) =y$? We need $F(x)= x^{1/n}I_{[0,1]}$, so $f(x)=\frac{1}{n}x^{1/n-1}I_{[0,1]}$, which is a Beta(1/n,1) density.

Let's check

> r<-replicate(100000, max(rbeta(4,1/4,1)))
> hist(r)

enter image description here


$F_{X_{(n)}}(x)=[F_X(x)]^n$, so for a standard uniform you need $F_X(x)=x^{1/n}$ for $0<x<1$ (and $0$ to the left and $1$ to the right of that interval), so $f_X(x)=\frac{1}{n}x^{\frac{1}{n}-1}$ on the unit interval and $0$ elsewhere.

It's a special case of the beta.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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