This is a continuation of the problem in Most powerful test given a sample from a variable with known but not common (I think) distribution

So having found the MP test for the simple hypothesis problem $H_0: \theta=0$ vs $H_1: \theta=2$, the next step asks if this test that was found is also UMP for the following problem:

$H_0: \theta<=0$ $H_1: \theta>0$

My intuition tells me that it should not be the UMP test since it was used only for $H_1: \theta=2$ and now our $H_1$ considers all values of $\theta>0$. However, the generalized test that was found in the first problem was to reject $H_0$ when:

$\frac{c(\theta)^nx^ne^{\frac{\theta}{2}\sum x_i^2}}{c(0)^nx^ne^{\frac{0}{2}\sum x_i^2}}>k$ which is also the same as

$\frac{(\frac{\theta}{e^{\theta/2}-1})^nx^ne^{\frac{2}{2}\sum x_i^2}}{2^nx^n}>k$

I recognize that in this form, every simple hypothesis test $H_1: \theta=\theta_1$ will be of the form $\sum x_i^2> t$ for some constant $t$ and for all $\theta_1>0$. Is this the UMP test that is being asked? Technically, it still is not the exact same test as the one when the value of $\theta$ in the alternative is fixed, right?


1 Answer 1


Note that your pdf $g(\theta)xe^{\theta \frac {x^{2}}{2}}$ is from the exponential family and has sufficient statistic $\sum X_i^2$.

Moreover, the pdf has monotone increasing likelihood. By Karlin-Rubin Theorem, the UMP test is to reject $H_0$ in favor of $H_1$ if $\sum X_i^2 \ge x_0$, where $x_0$ is chosen so that $P( \sum X_i^2 \ge x_0 ; \theta=0) = \alpha$.



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.