Say, we have a sequence of random variables $(X_i)_{1\leq i \leq n}$ that are i.i.d. and distributed according to a uniform distribution defined on interval $[0,1]$.

Why is $n-n\cdot\max\{X_1,\cdots,X_n\}$ distributed like an exponential distribution for real and positive $n$?

I know that $\lim_{n\rightarrow\infty}e(x)=(1+\frac{x}{n})^n$...

I also know that $\max\{X_1,\cdots,X_n\}=P(X_1)\cdot\cdots\cdot P(X_n)$ for independent variables...

I have:

\begin{align} F(x) & =P(n-n\cdot\max\{X_1,\cdots,X_n\}<x)=P(\max\{\ldots\} \\[10pt] & \geq1-x/n)=1-P(\max\{\ldots\}<1-x/n). \end{align}

Now, due to the cdf of $U(0,1)$ which is just $x$, this is equivalent to $1-(1-x/n)^n$.

But how do I proceed now? Or are there any mistakes so far?

  • $\begingroup$ Of course, $n-n\cdot\max\{X_1,\cdots,X_n\}$ is not distributed like an exponential distribution, for any positive integer $n$. $\endgroup$ – Did Jul 25 '18 at 11:40

Let $M_n = \max_{1\leqslant i\leqslant n}X_i$ for each positive integer $n$. For $0<t<1$ we have $$\{M_n \leqslant t\} = \bigcap_{i=1}^n \{X_i\leqslant t\},$$ so by independence it follows that $$ \mathbb P(M_n\leqslant t) = t^n. $$ Let $Y_n = n(1-M_n)$. Let $t>0$ and choose $N>1/t$. Then for $n\geqslant N$ we have \begin{align} \mathbb P(Y_n\leqslant t) &= \mathbb P(n(1-M_n)\leqslant t)\\ &=\mathbb P(1-M_n\leqslant t/n)\\ &=\mathbb P(M_n\geqslant 1-t/n)\\ &=1-\mathbb P(M_n\leqslant 1-t/n)\\ &=1 - (1-t/n)^n. \end{align} Since $$\lim_{n\to\infty}1 - (1-t/n)^n = 1 - e^{-t},$$ it follows that $Y_n$ converges in distribution to the exponential distribution with mean $1$.


You are right in your calculations. There are two points and you are done:

$1.$ From uniform to exponential you have the condition $0 \leq 1-x/n \leq 1$

$2.$ You have to consider $\lim n\rightarrow\infty$. Then you get $\exp(-x)$. Your formula about this is also correct.

So in conclusion you get an exponential distribution with parameter $+1$ and it is $exp(-1)$ safely enough it doesnt blow up.


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.