How to find cdf and pdf of an estimator

1.1k Views Asked by At

Consider a random variable X with a uniform distribution on $[0, a]$, with $(a > 0)$, where the value of a is unknown. To estimate a, we draw n independent random samples of $X$, denoted by $X_1,\ldots,X_n$, and we consider using :

$A_n$ = $\max(X_1,\ldots,X_n)$ to estimate the value of a.

1) What is the cdf and pdf of $An$?

2) Find $E[A_n]$ and $Var(A_n)$?

2

There are 2 best solutions below

2
On BEST ANSWER

First I point out how $X_i$ is distributed. In general the pdf of a uniform distributed variable has the pdf

$f_X(x)=\begin{cases} \frac1{u-l}, l<x<u \\0, \ \ \text{elsewhere}\end{cases}$

$u$ ist upper bound and $l$ ist the lower bound. In your cases $l=0$ and $u=a$

$f_X(x)=\begin{cases} \frac1{a}, 0<x<a \\0, \ \ \text{elsewhere}\end{cases}$

To get the cdf you calculate $\int_0^x f_X(t) \, dt=\int_0^x \frac1{a} \, dt= \ \frac{t}{a} \ \ \large|_0^x=\frac{x}{a}-\frac{0}{a}=\frac{x}{a}$

$F_X(x)=\begin{cases} 0, x\leq 0\\ \frac{x}{a}, 0<x<a \\1, \ \ x\geq a\end{cases}$


If $max(X_1, X_2,\dots, X_n)=x$, then each of the $n$ random variables, $X_i$, must be smaller or equal to $x$.

Since the samples are independent every sample has the same probability.

  • The probability that $X_1$ is smaller or equal than x is $\frac{x}a$, if $0<x<a$.

  • The probability that $X_2$ is smaller or equal than x is $\frac{x}a$, if $0<x<a$ as well.

  • The probability that $X_1$ and $X_2$ are smaller or equal than $x$ is $\frac{x}a\cdot \frac{x}a=\left( \frac{x}a \right)^2$

Now we can conclude what the probability is that every of the $n$ random variables are smaller or equal than $x$, if $0<x<a$.

I post it to make clear how the expected value has to be calculated.

$$P(A_n\leq x)=\left( \frac{x}a \right)^n$$

Differentiating w.r.t x gives $\left(\frac{1}a\right)^n\cdot n\cdot x ^{n-1}$

Thus $$E(A_n)=\int_0^a x\cdot \left(\frac{1}a\right)^n\cdot n\cdot x ^{n-1}\, dx=a\cdot \frac{n}{n+1}$$

2
On

Hint: for the cdf, $\mathbb P(A_n \le x) = \mathbb P(\text{all } X_i \le x)$. Pdf, mean and variance can be calculated from that.