Hypothesis testing, two equivalent definitions, mean value, unclear equivalence

62 Views Asked by At

We consider two definitions of the power function which equivalence is unclear to me.

(1) The power function of a test, denoted by $Q(\theta)$ is the probability of rejecting null hypothesis $H_0$ when $\theta\in \Theta$ is the true parameter value. The power function is then given by $$Q(\theta)=P_\theta(\cal R).$$

We have obviously here $Q(\theta_0)=\alpha$ and $Q(\theta_1)=1-\beta$ where $\alpha$ is the I type error and $\beta$ is the II type error.

(2) A function $\psi(.):\cal X^n\to [0,1]$ is called the critical function where $\psi(\mathbb{x})$ stands for the probability with which the null hypothesis $H_0$ is rejected when the data $\mathbb{X}=\mathbb{x}$ has been observed, $\mathbb{x}\in \cal X^n$.

Why now, we can rewrite the power function defined in (1) as follows

$$Q(\theta)=E_\theta\{\psi(\mathbb{X)}\}$$ for $\theta\in \Theta$ ? Where comes the expected value under the true $\theta$ from?

1

There are 1 best solutions below

4
On BEST ANSWER

$$P_\theta(\text{reject}) = E_\theta[1_{\text{reject}}] = E_\theta[E_\theta[1_{\text{reject}} \mid X]] = E_\theta[\psi(X)].$$