I'm solving the following exercise from Larry Wasserman's book "All of Statistics".

Let $X_1, \dots, X_n\sim \text{Unif}(0,\theta)$ and let $Y = \max{X_1,\dots,X_n}$. We want to test $H_0: \theta = 0.5$ versus $H_1:\theta > 0.5$. Suppose we decide to test this hypothesis by rejecting $H_0$ when $Y > c$.

  • a) Find the power function
  • b) What choice of $c$ will make the size of the test $0.05$?
  • c) In a sample of size $n=20$ with $Y=0.48$ what is the p-value? What conclusion about $H_0$ would you make?

The power function $\beta(\theta)$ is defined as $P_\theta(X\in R)$ where $R$ is the rejection area. So,

$$\beta(\theta) = P_\theta(Y > c) = 1 -P(\max_i X_i \le c) = 1- \prod_{i = 1}^n P_\theta(X_i\le c) = 1-c^n$$ if $\theta > c$ and $0$ otherwise. Is this correct?

For $b)$: the size is defined as $\alpha = \sup_{\theta\in\Theta_0}\beta(\theta)$. Since $\Theta_0$ is a singleton we have $$0.05 = P_{\theta_0}(Y > c) = 1-c^n\iff c = (1-0.05)^{\frac{1}{n}}$$ I think this is not correct since for $c)$: we have $n = 20$ which would imply $c \approx 0.9975$. This sounds too extreme. So where is my mistake? Moreover, for the $p$-value I would calculate:

$$p = P_{\theta_0}(\max_i X_i > 0.48) = (1-0.48^{20}) \approx 0.9999996$$ So that I would not reject $H_0$.


This sounds too extreme. So where is my mistake?

The issue is that under the null the density is $2$ for $0<x<0.5$

The cdf is $2x$, so $P(X_i<c) = 2c$, and $P(Y<c) = (2c)^n$ and $P(Y>c)=1-(2c)^n$.


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.