两独立样本率差异性检验¶
两样本率分别用 \(\hat{p}_1\) 和 \(\hat{p}_2\) 表示。
对于双侧检验，统计学假设如下：
\[
\begin{align}
H_0 &: p_1 = p_2 \\
H_1 &: p_1 \neq p_2
\end{align}
\]
对于左单侧检验，统计学假设如下：
\[
\begin{align}
H_0 &: p_1 \geqslant p_2 \\
H_1 &: p_1 \lt p_2
\end{align}
\]
对于右单侧检验，统计学假设如下：
\[
\begin{align}
H_0 &: p_1 \leqslant p_2\\
H_1 &: p_1 \gt p_2
\end{align}
\]
以下推导过程在边界条件 \(p_1 = p_2\) 下进行。
Z-Test Pooled¶
假设 \(\bar{p}\) 表示总体合并率：
\[
\bar{p} = \frac{n_1 p_1 + n_2 p_2}{n_1 + n_2}
\]
在 \(H_0\) 成立时，\(p_1 = p_2 = \bar{p}\)，两样本率差的方差可以用 \(\bar{p}\) 来表示：
\[
\operatorname{Var}(\hat{p}_1 - \hat{p}_2) = \frac{p_1(1-p_1)}{n_1} + \frac{p_2(1-p_2)}{n_2} = \bar{p}(1-\bar{p}) \left(\frac{1}{n_1} + \frac{1}{n_2}\right)
\]
可构建 \(z\) 统计量：
\[
z = \frac{\hat{p}_1 - \hat{p}_2}{\sqrt{\bar{p}(1-\bar{p}) \left(\frac{1}{n_1} + \frac{1}{n_2}\right)}} \sim N(0,1)
\]
在 \(H_1\) 成立时，两样本率差的方差如下：
\[
\operatorname{Var}(\hat{p}_1 - \hat{p}_2) = \frac{p_1(1-p_1)}{n_1} + \frac{p_2(1-p_2)}{n_2}
\]
可构建 \(z'\) 统计量：
\[
z' = \frac{\hat{p}_1 - \hat{p}_2}{\sqrt{\bar{p}(1-\bar{p}) \left(\frac{1}{n_1} + \frac{1}{n_2}\right)}}
\sim N\left(\frac{p_1-p_2}{\sqrt{\bar{p}(1-\bar{p}\left(\frac{1}{n_1} + \frac{1}{n_2}\right))}},
\frac{\frac{p_1(1-p_1)}{n_1} + \frac{p_2(1-p_2)}{n_2}}{\bar{p}(1-\bar{p})\left(\frac{1}{n_1} + \frac{1}{n_2}\right)}
\right)
\]
\[
\begin{align}
\text{Power}
& = P\left(z' > z_{1-\alpha/2}\right) + P\left(z' < z_{\alpha/2}\right) \\
& = 1 - \Phi\left(\frac{z_{1-\alpha/2} \sqrt{\bar{p}(1-\bar{p}) \left(\frac{1}{n_1} + \frac{1}{n_2}\right)} - \left(p_1-p_2\right)}
{\sqrt{\frac{p_1(1-p_1)}{n_1} + \frac{p_2(1-p_2)}{n_2}}}
\right)
+ \Phi\left(\frac{z_{\alpha/2} \sqrt{\bar{p}(1-\bar{p}) \left(\frac{1}{n_1} + \frac{1}{n_2}\right)} - \left(p_1-p_2\right)}
{\sqrt{\frac{p_1(1-p_1)}{n_1} + \frac{p_2(1-p_2)}{n_2}}}
\right)
\end{align}
\]
\[
\begin{align}
\text{Power} = P\left(z' < z_{\alpha}\right)
= \Phi\left(\frac{z_{\alpha} \sqrt{\bar{p}(1-\bar{p}) \left(\frac{1}{n_1} + \frac{1}{n_2}\right)} - \left(p_1-p_2\right)}
{\sqrt{\frac{p_1(1-p_1)}{n_1} + \frac{p_2(1-p_2)}{n_2}}}
\right)
\end{align}
\]
\[
\begin{align}
\text{Power} = P\left(z' > z_{1-\alpha}\right)
= 1 - \Phi\left(\frac{z_{1-\alpha} \sqrt{\bar{p}(1-\bar{p}) \left(\frac{1}{n_1} + \frac{1}{n_2}\right)} - \left(p_1-p_2\right)}
{\sqrt{\frac{p_1(1-p_1)}{n_1} + \frac{p_2(1-p_2)}{n_2}}}
\right)
\end{align}
\]
单侧检验样本量公式推导
根据标准正态分布分位数的定义：
\[
\frac{z_{1-\alpha} \sqrt{\bar{p}(1-\bar{p}) \left(\frac{1}{n_1} + \frac{1}{n_2}\right)} \pm (p_1-p_2)}{\sqrt{\frac{p_1(1-p_1)}{n_1} + \frac{p_2(1-p_2)}{n_2}}} = z_\beta
\]
设 \(n_1 = kn_2\)，由上式可解出：
\[
n_2 = \frac{\left(z_{1-\alpha} \sqrt{\bar{p}(1-\bar{p})(1/k+1)} + z_{1-\beta} \sqrt{p_1(1-p_1)/k + p_2(1-p_2)}\right)^2}{(p_1-p_2)^2}
\]
\[
n_1 = k n_2
\]
Z-Test Pooled with Continuity Correction¶
在 Z-Test Pooled 的基础上，添加校正项。
定义：
\[
c = \frac{1}{2}\left(\frac{1}{n_1}+\frac{1}{n_2}\right)
\]
校正项的符号由检验方向决定，左侧检验时，校正项为 \(+c\)，右侧检验时，校正项为 \(-c\)。
在 \(H_0\) 成立时，可构建 \(z\) 统计量：
\[
z = \frac{\hat{p}_1 - \hat{p}_2 \pm c}{\sqrt{\bar{p}(1-\bar{p}) \left(\frac{1}{n_1} + \frac{1}{n_2}\right)}} \sim N(0,1)
\]
在 \(H_1\) 成立时，可构建 \(z'\) 统计量：
\[
z' = \frac{\hat{p}_1 - \hat{p}_2 \pm c}{\sqrt{\bar{p}(1-\bar{p}) \left(\frac{1}{n_1} + \frac{1}{n_2}\right)}}
\sim N\left(\frac{p_1 - p_2 \pm c}{\sqrt{\bar{p}(1-\bar{p}\left(\frac{1}{n_1} + \frac{1}{n_2}\right))}},
\frac{\frac{p_1(1-p_1)}{n_1} + \frac{p_2(1-p_2)}{n_2}}{\bar{p}(1-\bar{p})\left(\frac{1}{n_1} + \frac{1}{n_2}\right)}
\right)
\]
\[
\begin{align}
\text{Power} & = P\left(z' > z_{1-\alpha/2}\right) + P\left(z' < z_{\alpha/2}\right) \\
& = \begin{aligned}[t]
1 & - \Phi\left(\frac{z_{1-\alpha/2} \sqrt{\bar{p}(1-\bar{p}) \left(\frac{1}{n_1} + \frac{1}{n_2}\right)} - (p_1-p_2-c)}
{\sqrt{\frac{p_1(1-p_1)}{n_1} + \frac{p_2(1-p_2)}{n_2}}}
\right) \\
& + \Phi\left(\frac{z_{\alpha/2} \sqrt{\bar{p}(1-\bar{p}) \left(\frac{1}{n_1} + \frac{1}{n_2}\right)} - (p_1-p_2+c)}
{\sqrt{\frac{p_1(1-p_1)}{n_1} + \frac{p_2(1-p_2)}{n_2}}}
\right)
\end{aligned}
\end{align}
\]
\[
\begin{align}
\text{Power} = P\left(z' < z_{\alpha}\right)
= \Phi\left(\frac{z_{\alpha} \sqrt{\bar{p}(1-\bar{p}) \left(\frac{1}{n_1} + \frac{1}{n_2}\right)} - (p_1-p_2+c)}
{\sqrt{\frac{p_1(1-p_1)}{n_1} + \frac{p_2(1-p_2)}{n_2}}}
\right)
\end{align}
\]
\[
\begin{align}
\text{Power} = P\left(z' > z_{1-\alpha}\right)
= 1 - \Phi\left(\frac{z_{1-\alpha} \sqrt{\bar{p}(1-\bar{p}) \left(\frac{1}{n_1} + \frac{1}{n_2}\right)} - (p_1-p_2-c)}
{\sqrt{\frac{p_1(1-p_1)}{n_1} + \frac{p_2(1-p_2)}{n_2}}}
\right)
\end{align}
\]
Z-Test Unpooled¶
在 \(H_0\) 成立时，可构建 \(z\) 统计量：
\[
z = \frac{\hat{p}_1 - \hat{p}_2}{\sqrt{\frac{p_1(1-p_1)}{n_1} + \frac{p_2(1-p_2)}{n_2}}} \sim N(0,1)
\]
在 \(H_1\) 成立时，可构建 \(z'\) 统计量：
\[
z' = \frac{\hat{p}_1 - \hat{p}_2}{\sqrt{\frac{p_1(1-p_1)}{n_1} + \frac{p_2(1-p_2)}{n_2}}} \sim N\left(\frac{p_1 - p_2}{\sqrt{\frac{p_1(1-p_1)}{n_1} + \frac{p_2(1-p_2)}{n_2}}}, \ 1\right)
\]
\[
\begin{align}
\text{Power}
& = P\left(z' > z_{1-\alpha/2}\right) + P\left(z' < z_{\alpha/2}\right) \\
& = 1 - \Phi\left(z_{1-\alpha/2} - \frac{p_1 - p_2}{\sqrt{\frac{p_1(1-p_1)}{n_1} + \frac{p_2(1-p_2)}{n_2}}}\right)
+ \Phi\left(z_{\alpha/2} - \frac{p_1 - p_2}{\sqrt{\frac{p_1(1-p_1)}{n_1} + \frac{p_2(1-p_2)}{n_2}}}\right)
\end{align}
\]
\[
\text{Power}
= P\left(z' < z_{\alpha}\right)
= \Phi\left(z_{\alpha} - \frac{p_1 - p_2}{\sqrt{\frac{p_1(1-p_1)}{n_1} + \frac{p_2(1-p_2)}{n_2}}}\right)
\]
\[
\text{Power}
= P\left(z' > z_{1-\alpha}\right)
= 1 - \Phi\left(z_{1-\alpha} - \frac{p_1 - p_2}{\sqrt{\frac{p_1(1-p_1)}{n_1} + \frac{p_2(1-p_2)}{n_2}}}\right)
\]
单侧检验样本量公式推导
根据标准正态分布分位数的定义：
\[
z_{1-\alpha} \pm \frac{p_1 - p_2}{\sqrt{\frac{p_1(1-p_1)}{n_1} + \frac{p_2(1-p_2)}{n_2}}} = z_\beta
\]
设 \(n_1 = kn_2\)，由上式可解出：
\[
n_2 = \frac{\left(z_{1-\alpha} + z_{1-\beta}\right)^2 \left[ p_1(1-p_1)/k + p_2(1-p_2) \right]}{(p_1-p_2)^2}
\]
\[
n_1 = k n_2
\]
Z-Test Unpooled with Continuity Correction¶
在 Z-Test Unpooled 的基础上，添加校正项。
定义：
\[
c = \frac{1}{2}\left(\frac{1}{n_1}+\frac{1}{n_2}\right)
\]
校正项的符号由检验方向决定，左侧检验时，校正项为 \(+c\)，右侧检验时，校正项为 \(-c\)。
在 \(H_0\) 成立时，可构建 \(z\) 统计量：
\[
z = \frac{\hat{p}_1 - \hat{p}_2 \pm c}{\sqrt{\frac{p_1(1-p_1)}{n_1} + \frac{p_2(1-p_2)}{n_2}}} \sim N(0,1)
\]
在 \(H_1\) 成立时，可构建 \(z'\) 统计量：
\[
z' = \frac{\hat{p}_1 - \hat{p}_2 \pm c}{\sqrt{\frac{p_1(1-p_1)}{n_1} + \frac{p_2(1-p_2)}{n_2}}} \sim N\left(\frac{p_1 - p_2 \pm c}{\sqrt{\frac{p_1(1-p_1)}{n_1} + \frac{p_2(1-p_2)}{n_2}}}, \ 1\right)
\]
\[
\begin{align}
\text{Power}
& = P\left(z' > z_{1-\alpha/2}\right) + P\left(z' < z_{\alpha/2}\right) \\
& = 1 - \Phi\left(z_{1-\alpha/2} - \frac{p_1-p_2-c}{\sqrt{\frac{p_1(1-p_1)}{n_1} + \frac{p_2(1-p_2)}{n_2}}}\right)
+ \Phi\left(z_{\alpha/2} - \frac{p_1-p_2+c}{\sqrt{\frac{p_1(1-p_1)}{n_1} + \frac{p_2(1-p_2)}{n_2}}}\right)
\end{align}
\]
\[
\text{Power}
= P\left(z' < z_{\alpha}\right)
= \Phi\left(z_{\alpha} - \frac{p_1-p_2+c}{\sqrt{\frac{p_1(1-p_1)}{n_1} + \frac{p_2(1-p_2)}{n_2}}}\right)
\]
\[
\text{Power}
= P\left(z' > z_{1-\alpha}\right)
= 1 - \Phi\left(z_{1-\alpha} - \frac{p_1-p_2-c}{\sqrt{\frac{p_1(1-p_1)}{n_1} + \frac{p_2(1-p_2)}{n_2}}}\right)
\]