Let $X_1, X_2, ... , X_n$ be a random sample from the uniform distribution over $[0, \theta]$. Suppose we wish to test $H_0 : \theta = 5$ versus $H_A : \theta < 5$ at significance level $\alpha = 0.05$. Use the LRT to find the critical region.

Here is my attempt:

\begin{align*} T = \frac{L(\theta_0 = 5)}{L(\theta_A < 5)} & = \frac{\Pi_{i = 1}^n \frac{1}{b-a}}{\Pi_{i = 1}^n \frac{1}{b-a}} \\ & = \frac{\frac{1}{(5-a)^n}}{\frac{1}{(\theta_A-a)^n}} \\ & = \frac{(\theta_A-a)^n}{(5-a)^n} \end{align*}

However, I am not sure how to proceed next. I know that we reject the null hypothesis if $T < c$ where $c$ is some constant. In my book they give an example where they now take $\ln$ on both sides and if I do so I get $$ \ln(T) = - n \ln(\theta_A - a) + n \ln(5 - a)) < \ln(c) $$

but what am I supposed to solve for?

  • $\begingroup$ Is this for a homework assignment? $\endgroup$
    – AdamO
    Mar 23, 2021 at 15:55
  • $\begingroup$ No, it is just homework that I have to prepare for my class on Thursday (we go through all of the exercises then but I just want to understand everything before class) $\endgroup$
    – Mathias
    Mar 23, 2021 at 15:57

1 Answer 1


The likelihood ratio is $$ \lambda(x) = \frac{\sup_{\theta\in\Theta_0} L(\theta\mid x)}{\sup_{\theta\in\Theta} L(\theta\mid x)}. $$ In this case $\Theta_0 = \{5\}$, $\Theta = [0,5]$, and your likelihood is $$ L(\theta\mid x) = \theta^{-n}\mathbf 1_{x_\max \leq \theta }. $$ On $[0,5]$ $\theta^{-n}$ is strictly decreasing so we want the smallest $\theta$ possible such that we don't violate the support, and this leads to $\sup_{\theta\in\Theta} L(\theta\mid x) = L(x_\max \mid x)$. This means our likelihood ratio is $$ \lambda(x) = \frac{L(5 \mid x)}{L(x_\max \mid x)} = \left(\frac{5}{x_\max}\right)^{-n}. $$ $x_\max$ is a sufficient statistic for $\theta$ here so it's nice to see that our LRT only depends on this sufficient statistic. You got this part already but I think your notation was a little unclear since $a$ and $b$ weren't really defined.

For the rejection region, we can get there by thinking about what we're trying to do. We'll reject the null hypothesis if $\lambda(x)$ is small, because this means there's a much higher likelihood outside of $\Theta_0$ than in it. To get our threshold then we know that we want the probability of a rejection under $H_0$ to be $\alpha$, so this means we need to solve $$ P(\lambda(X) \leq c \mid \theta \in \Theta_0) = \alpha $$ for $c$. The null hypothesis is simple here so things are a little easier: $$ P\left(\lambda(X) \leq c \mid \theta \in \Theta_0\right) = P\left(\left(\frac{5}{X_\max}\right)^{-n} \leq c\right) = \alpha. $$ At this point can you finish the problem?

  • $\begingroup$ where from the question stem do you find $\Theta = [0,5] $ ? $\endgroup$ Jul 10, 2022 at 16:48

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

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