2
$\begingroup$

I'm going through Statistical Inference by Casella & Berger, and on page 419, in the intro section of interval estimation there is the following example (note: most of the text was left out as it's irrelevant to my question):

Let $X_1,...,X_n$ be a random sample from a $\text{Uniform}(0, \theta)$ population, and let $Y = \max{\{X_1,...,X_n\}}$, ... Note that $\theta$ is necessarily larger than $y$.

My question is why is $\theta$ necessarily larger than $y$.

$\endgroup$
  • 5
    $\begingroup$ What is your understanding of a "uniform(0, theta) population"? $\endgroup$ – whuber Aug 17 '18 at 15:20
  • 2
    $\begingroup$ In a continuous distribution, such as $\text{Uniform}(0,\theta)$ what is $P(X)=c$ for some specific real value $c$? And for that distribution what is $P(X)>\theta$? $\endgroup$ – Alexis Aug 17 '18 at 15:29
  • 2
    $\begingroup$ @Alexis My current thinking is because X is distributed as a continuous uniform, then $P(X=x)=0$, so $P(X_{(n)}=\theta)=0$. Perhaps I'm still somewhat confused when thinking about zero probability events and realized values of a continuous random variable. $\endgroup$ – Zian Aug 17 '18 at 16:05
  • 2
    $\begingroup$ The relevant probability is $\Pr(Y \gt \theta).$ Can you compute that? $\endgroup$ – whuber Aug 17 '18 at 16:31
  • $\begingroup$ I think it is is better to say '$Y$ is smaller than or equal to $\theta$' instead of '$Y$ is smaller than $\theta$' or equivalent '$\theta $ is bigger than $Y$'. Do you get why each of the $X_i $ should be smaller than or equal to $\theta $? $\endgroup$ – Martijn Weterings Aug 17 '18 at 18:09
2
$\begingroup$

Your thinking is correct. Since the probability if $\theta$ being bigger than $Y$ is equal to 1:

$$ \begin{split} P(Y < \theta) &= P(X_1 < \theta \ \cap \ ... \ \cap \ X_n < \theta)\\ &= P(X_1 < \theta)\ ...\ P(X_n < \theta)\\ &= [F(\theta) - P(X_1 = \theta)] \ ... \ [F(\theta) - P(X_n = \theta)] \\ &= [F(\theta) - 0] \ ... \ [F(\theta) - 0]\\ &= 1 \cdot 1 \cdot \ ... \ \cdot 1\\ &= 1 \end{split} $$

Where $F(x)$ is the CDF of the $\text{Uniform}(0, \theta)$ population and we are assuming the $X_1 ... X_n$ are independent.

$ F(x)= \begin{cases} 0 & \text{for }x < 0 \\ \frac{x}{\theta} & \text{for }0 \le x \le \theta \\ 1 & \text{for }x > \theta \end{cases}$

$\endgroup$

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.