I've got a problem, with a solution, in my introduction to mathematical statistics book and I just don't get how they got there. There are follow up questions so I'd like to get insight at how they got to the answer.

The problem: Let $X_1,\ldots,X_n$ be independent random variables with the uniform distribution on the interval $[0,1]$. Determine the expectation and variance of $Y=\max(X_1,\ldots,X_n)$. Hint: Deduce the density of Y from the distriution function $P(Y \leq y)$ of $Y$, which can be determined using the distribution functions of $X_1, \ldots, X_n$.

The solution: $\operatorname E[Y]=\frac{n}{n+1}$, $\operatorname{var}[Y]=\frac{n}{n+2}+\left(\frac{n}{n+1}\right)^2$.

Now I know that a uniformly distributed random variable on $[0,1]$ has the following probability distribution, expectation and variance;

$F(x)=x$ for $x \in [0,1]$, $\operatorname E[X]=\frac{1}{2}$ and $\operatorname{var}[X]=\frac{1}{12}$.

  • 3
    $\begingroup$ Have you tried to start with the hint? For $y\in[0,1]$, $$\mathbb{P}\{ Y \leq y \} = \mathbb{P}\{ \max(X_1,\dots,X_n) \leq y \} = \mathbb{P}\{ \forall 1\leq i\leq n,\ X_i \leq y \} = \mathbb{P}(\{ X_1 \leq y \}\cap\{ X_2 \leq y \}\cap\dots\cap\{ X_n \leq y \}) = \dots$$ $\endgroup$ – Clement C. Sep 4 '18 at 19:55
  • $\begingroup$ Thanks for your comment. I would say all those individual probabilities are equal to $y$. Thus $ P(\{X_{1} \leq y\} \cap ... \cap \{X_{n} \leq y\}) = y^n $. Can you confirm? $\endgroup$ – Mathbeginner Sep 4 '18 at 20:03
  • $\begingroup$ Incidentally, the solution for the variance is wrong: the "+" should be a "-". (Sanity check: when $n\to\infty$, the variance would converge to $2$.... you cannot have a variance greater than $1$ for a random variable in $[0,1]$!) $\endgroup$ – Clement C. Sep 4 '18 at 20:03
  • $\begingroup$ Yes, this is $y^n$ -- the reason is that the $X_i$'s are independent, so the probability of the intersection is equal to the product of the probabilities. Now, you have the cdf of $Y$. Use it to derive the expectation and variance. $\endgroup$ – Clement C. Sep 4 '18 at 20:04
  • 1
    $\begingroup$ No, it is not a typo (see the link at the end of the comment). $\endgroup$ – Clement C. Sep 4 '18 at 20:37

Here is an alternative approach which doesn't require any knowledge of the Beta distribution.


$$Y=\max\{X_1,\ldots,X_n\} \leq y \iff \forall j=1,\ldots,n: X_j \leq y$$

it follows from the independence of the random variables $(X_j)_{j \leq n}$ that

$$\mathbb{P}(Y \leq y) = \mathbb{P} \left( \bigcap_{j=1}^n \{X_j \leq y\} \right) = \prod_{j=1}^n \mathbb{P}(X_j \leq y) = y^n \tag{1}$$

for $y \in (0,1)$. Thus the distribution function $F$ of $Y$ satisfies

$$F(y) = \begin{cases} 0, & y \leq 0, \\ y^n, & y \in (0,1), \\ 1, & y \geq 1. \end{cases}$$

Approach 1: Differentiating $F$ we find that $Y$ has a density $p$ with respect to Lebesgue measure, $$p(y) = n y^{n-1} 1_{(0,1)}(y).$$ Thus $$\mathbb{E}(Y)= \int y p(y) \, dy =n \int_0^1 y^n \, dy = \frac{n}{n+1}$$ and $$\mathbb{E}(Y^2) = \int y^2 p(y) \, dy = n \int_0^1 y^{n+1} \, dy = \frac{n}{n+2}.$$ Thus, $$\begin{align*} \text{var}(Y) = \mathbb{E}(Y^2)-(\mathbb{E}(Y))^2 = \frac{n}{n+2} - \frac{n^2}{(n+1)^2} &= \frac{n(n+1)^2 - n^2 (n+2)}{(n+1)^2 (n+2)} \\ &= \frac{n}{(n+1)^2 (n+2)}. \end{align*}$$

Approach 2: For any non-negative random variable $Z$ it holds that

$$\mathbb{E}(Z) = \int_0^{\infty} \mathbb{P}(Z > z) \, dz = \int_0^{\infty} (1-\mathbb{P}(Z \leq z)) \, dz, \tag{2}$$

see e.g. this question for details. For $Z:=Y$ we obtain from $(1)$ that

$$\mathbb{E}(Y) = \int_0^{1} (1-y^n) \, dy= 1 - \frac{1}{n+1} = \frac{n}{n+1}.$$


$$\mathbb{E}(Y^2) \stackrel{(2)}{=} \int_0^{\infty} (1-\mathbb{P}(Y^2 \leq y)) \, dy = \int_0^{\infty} (1-\mathbb{P}(Y \leq \sqrt{y})) \, dy,$$

and so by $(1)$

$$\mathbb{E}(Y^2) = \int_0^1 (1- y^{n/2}) \, dy = 1- \frac{1}{\frac{n}{2}+1} = \frac{n}{n+2}.$$

As in the first approach, we thus find

$$\begin{align*} \text{var}(Y) = \mathbb{E}(Y^2)-(\mathbb{E}(Y))^2 = \frac{n}{n+2} - \frac{n^2}{(n+1)^2} = \frac{n}{(n+1)^2 (n+2)}. \end{align*}$$

  • $\begingroup$ No beta. Very natural. Brilliant! $\endgroup$ – BCLC Sep 5 '18 at 5:55
  • $\begingroup$ For someone not familiar with the beta distribution such as myself, a very nice way. Thanks :) $\endgroup$ – Mathbeginner Sep 6 '18 at 6:52
  • $\begingroup$ @Mathbeginner You are welcome. If you like the answer, you can upvote it by clicking on the up arrow next to it. $\endgroup$ – saz Sep 6 '18 at 7:06

Notice that $$\mathbb{P}(Y \leq y) = \mathbb{P}(\max(X_1, \dots, X_n) \leq y)\text{.}$$ Now, note that if the largest of $n$ numbers is less than $y$, it follows that each of the $n$ numbers must be less than $y$ as well.

So, $$\mathbb{P}(\max(X_1, \dots, X_n) \leq y) = \mathbb{P}(X_1 \leq y \cap X_2 \leq y \cap\cdots\cap X_n \leq y)\text{.}$$ By independence, we have $$\mathbb{P}(X_1 \leq y \cap X_2 \leq y \cap\cdots\cap X_n \leq y) =\mathbb{P}(X_1 \leq y) \mathbb{P}(X_2 \leq y)\cdots\mathbb{P}(X_n \leq y)=y^n$$ for $y \in [0, 1]$. Hence, $$F_{Y}(y)=\mathbb{P}(Y \leq y) = \begin{cases} 0, & y < 0 \\ y^n, & y \in [0, 1] \\ 1, & y > 1\text{.} \end{cases}$$ The PDF of $Y$ is thus $$f_{Y}(y)=F^{\prime}_{Y}(y) = \begin{cases} ny^{n-1}, & y \in [0, 1] \\ 0, & \text{otherwise.} \end{cases}$$

Thus, $Y$ follows a Beta distribution with $\alpha = n$ and $\beta = 1$. Hence, the mean is $$\mathbb{E}[Y]=\dfrac{\alpha}{\alpha+\beta}=\dfrac{n}{n+1}$$ and the variance is $$\text{Var}(Y) = \dfrac{\alpha\beta}{(\alpha+\beta)^2(\alpha+\beta+1)} = \dfrac{n}{(n+1)^2(n+2)}\text{.}$$


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.