1
$\begingroup$

I am trying to understand the proof of Neyman Pearson Lemma as Uniformly Most Powerful test from here (Page 3).

It says the following:

Let $H_0: \theta = \theta_0$ and $H_a: \theta = \theta_1$. Also, let the test-statistic of NP be $\Lambda(x)$ and that of a different test be $K(x)$. For a given $\alpha$ level, let $c$ and $d$ be the critical values for $\Lambda(x)$ and $K(x)$ respectively.

Let \begin{align} & \text{A = [the likelihood-ratio test rejects $H_0$ and the other test does not],} \\ & \text{B = [the other test rejects $H_0$ and the likelihood-ratio test does not],} \\ \text{and } & \text{C = [both tests reject $H_0$]}\end{align}

It offers an simple proof as the following:

\begin{align} & \text{Pr}(A|H_1) = \int_A f(x | \theta = \theta_1) dx \gt \int_A c f(x | \theta = \theta_0) dx = c\text{Pr}(A|H_0) \\ \overset{?}{= } & \, \, c\text{Pr}(B|H_0) = \int_B c f(x | \theta = \theta_0) dx \gt \int_B f(x | \theta = \theta_1) dx = \text{Pr}(B|H_1)\end{align}

I am unable to see why the questionmarked equality is valid.

$\endgroup$

1 Answer 1

1
$\begingroup$

Both tests have the same significance level; assuming that they both attain that level exactly:

$P(\text{LRT rejects}|H_0)=P(\text{other test rejects}|H_0)=\alpha$

$\therefore P(\text{LRT rejects & other doesn't}|H_0)+P(\text{both reject}|H_0)\\\qquad=P(\text{other test rejects & LRT doesn't}|H_0)+P(\text{both reject}|H_0)$

$\text{or } P(\text{LRT rejects & other doesn't}|H_0)=P(\text{other test rejects & LRT doesn't}|H_0)$

$\endgroup$

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.