应用概率统计 陈魁 清华大学出版社 统计部分 期末考点

一、参数估计

1.点估计

1.1矩法(用样本矩作为相应的总体矩的估计量)

思路：解方程（组）
$$E(X^k)=\frac{1}{n}\sum _{i=1}^n{X}_i^k.$$

1.2极大似然法(使得样本落在观察值的领域里的概率(似然函数)最大)

1. 写出似然函数
   离散型
   $$L(x_1,x_2,\cdots ,x_n;\theta )=\prod _{i=1}^{n}p(X=x_i;\theta ),$$
   连续型
   $$L(x_1,x_2,\cdots ,x_n;\theta )=\prod _{i=1}^{n}f(x_i;\theta )\mathrm{d}{x}_i,$$
2. 写出$\ln L;$
3. 求出$\frac{\mathrm{d}\ln L}{\mathrm{d}\theta }$，并令其为0，解出$\theta$，即为$\hat{\theta }.$

2.区间估计

正态总体参数的区间估计(应用概率统计 陈魁 清华大学出版社)

二、假设检验

	原假设$H_0$	$H_0$下的检验统计量及分布	备择假设$H_1$	$H_0$的拒绝域
1	$$\begin{gathered} \mu ={\mu }_0 \\ ({\sigma }^2已知) \end{gathered}$$	$$U=\frac{\overline{X}-{\mu }_0}{\sigma /\sqrt{n}}\sim N(0,1)$$	$$\begin{gathered} \mu \ne {\mu }_0 \\ \mu >{\mu }_0 \\ \mu <{\mu }_0 \end{gathered}$$	$$\begin{gathered} |U|>z_{\frac{\alpha }{2}} \\ U>z_{\alpha } \\ U<-z_{\alpha } \end{gathered}$$
2	$$\begin{gathered} \mu ={\mu }_0 \\ ({\sigma }^2未知) \end{gathered}$$	$$T=\frac{\overline{X}-{\mu }_0}{S/\sqrt{n}}\sim t(n-1)$$	$$\begin{gathered} \mu \ne {\mu }_0 \\ \mu >{\mu }_0 \\ \mu <{\mu }_0 \end{gathered}$$	$$\begin{gathered} |T|>t_{\frac{\alpha }{2}}(n-1) \\ T>t_{\alpha }(n-1) \\ T<-t_{\alpha }(n-1) \end{gathered}$$
3	$$\begin{gathered} {\mu }_1-{\mu }_2=\delta \\ ({\sigma }_1^2,{\sigma }_2^2均已知) \end{gathered}$$	$$U=\frac{\overline{X}-\overline{Y}-\delta }{\sqrt{\frac{{\sigma }_1^2}{n_1}+\frac{{\sigma }_2^2}{n_2}}}\sim N(0,1)$$	$$\begin{gathered} {\mu }_1-{\mu }_2\ne \delta \\ {\mu }_1-{\mu }_2>\delta \\ {\mu }_1-{\mu }_2<\delta \end{gathered}$$	$$\begin{gathered} |U|>z_{\frac{\alpha }{2}} \\ U>z_{\alpha } \\ U<-z_{\alpha } \end{gathered}$$
4	$$\begin{gathered} {\mu }_1-{\mu }_2=\delta \\ \left(\begin{array}{c}{\sigma }_1^2,{\sigma }_2^2均未知\\ {\sigma }_1^2={\sigma }_2^2={\sigma }^2\end{array}\right) \end{gathered}$$	$$\begin{gathered} T=\frac{\overline{X}-\overline{Y}-\delta }{S_w\sqrt{\frac{1}{n_1}+\frac{1}{n_2}}}\sim t(n_1+n_2-2), \\ {S}_w^2=\frac{(n_1-1)S_1^2+(n_2-1)S_2^2}{n_1+n_2-2)} \end{gathered}$$	$$\begin{gathered} {\mu }_1-{\mu }_2\ne \delta \\ {\mu }_1-{\mu }_2>\delta \\ {\mu }_1-{\mu }_2<\delta \end{gathered}$$	$$\begin{gathered} |T|>t_{\frac{\alpha }{2}}(n_1+n_2-2) \\ T>t_{\alpha }(n_1+n_2-2) \\ T<-t_{\alpha }(n_1+n_2-2 \end{gathered}$$
5	$$\begin{gathered} {\mu }_1-{\mu }_2=\delta \\ \left(\begin{array}{c}{\sigma }_1^2,{\sigma }_2^2均未知\\ n_1,n_2很大\end{array}\right) \end{gathered}$$	$$U=\frac{\overline{X}-\overline{Y}-\delta }{\sqrt{\frac{S_1^2}{n_1}+\frac{S_2^2}{n_2}}}\sim N(0,1)$$	$$\begin{gathered} {\mu }_1-{\mu }_2\ne \delta \\ {\mu }_1-{\mu }_2>\delta \\ {\mu }_1-{\mu }_2<\delta \end{gathered}$$	$$\begin{gathered} |U|>z_{\frac{\alpha }{2}} \\ U>z_{\alpha } \\ U<-z_{\alpha } \end{gathered}$$
6	$$\begin{gathered} {\sigma }^2={\sigma }_0^2 \\ (\mu 已知) \end{gathered}$$	$${\kappa }^2=\sum _{i=1}^n(\frac{X_i-\mu }{{\sigma }_0}{)}^2\sim {\chi }^2(n)$$	$$\begin{gathered} {\sigma }^2\ne {\sigma }_0^2 \\ {\sigma }^2>{\sigma }_0^2 \\ {\sigma }^2<{\sigma }_0^2 \end{gathered}$$	$$\begin{gathered} {\kappa }^2>{\chi }_{\frac{\alpha }{2}}^2(n)或{\kappa }^2<{\chi }_{1-\frac{\alpha }{2}}^2(n) \\ {\kappa }^2>{\chi }_{\alpha }^2(n) \\ {\kappa }^2<{\chi }_{1-\alpha }^2(n) \end{gathered}$$
7	$$\begin{gathered} {\sigma }^2={\sigma }_0^2 \\ (\mu 未知) \end{gathered}$$	$${\kappa }^2=\frac{(n-1)S^2}{{\sigma }_0^2}\sim {\chi }^2(n-1)$$	$$\begin{gathered} {\sigma }^2\ne {\sigma }_0^2 \\ {\sigma }^2>{\sigma }_0^2 \\ {\sigma }^2<{\sigma }_0^2 \end{gathered}$$	$$\begin{gathered} {\kappa }^2>{\chi }_{\frac{\alpha }{2}}^2(n-1)或{\kappa }^2<{\chi }_{1-\frac{\alpha }{2}}^2(n-1) \\ {\kappa }^2>{\chi }_{\alpha }^2(n-1) \\ {\kappa }^2<{\chi }_{1-\alpha }^2(n-1) \end{gathered}$$
8	$$\begin{gathered} {\sigma }_1^2={\sigma }_2^2 \\ ({\mu }_1,{\mu }_2均已知) \end{gathered}$$	$$F=\frac{\sum _{i=1}^{n_1}(\frac{X_i-{\mu }_1}{{\sigma }_1}{)}^2/n1}{\sum _{i=1}^{n_2}(\frac{X_i-{\mu }_2}{{\sigma }_2}{)}^2/n2}\sim F(n_1,n_2)$$	$$\begin{gathered} {\sigma }_1^2\ne {\sigma }_2^2 \\ {\sigma }_1^2>{\sigma }_2^2 \\ {\sigma }_1^2<{\sigma }_2^2 \end{gathered}$$	$$\begin{gathered} F>F_{\frac{\alpha }{2}}(n_1,n_2)或F<F_{1-\frac{\alpha }{2}}(n_1,n_2) \\ F>F_{\alpha }(n_1,n_2) \\ F<F_{1-\alpha }(n_1,n_2) \end{gathered}$$
9	$$\begin{gathered} {\sigma }_1^2={\sigma }_2^2 \\ ({\mu }_1,{\mu }_2均未知) \end{gathered}$$	$$F=\frac{S_1^2}{S_2^2}\sim F(n_1-1,n_2-1)$$	$$\begin{gathered} {\sigma }_1^2\ne {\sigma }_2^2 \\ {\sigma }_1^2>{\sigma }_2^2 \\ {\sigma }_1^2<{\sigma }_2^2 \end{gathered}$$	$$\begin{gathered} F>F_{\frac{\alpha }{2}}(n_1-1,n_2-1)或F<F_{1-\frac{\alpha }{2}}(n_1-1,n_2-1) \\ F>F_{\alpha }(n_1-1,n_2-1) \\ F<F_{1-\alpha }(n_1-1,n_2-1) \end{gathered}$$

三、方差分析

方差分析(应用概率统计 陈魁 清华大学出版社)

四、回归分析

1.一元(正态)线性回归(最小二乘估计)

• 样本$(x_1,y_1),(x_2,y_2),\cdots ,(x_i,y_i),\cdots ,(x_n,y_n);$
• 模型$$\begin{array}{lr}{y}_i=a+b{x}_i+{\epsilon }_i,& i=1,2,\cdots ,n\\ {\epsilon }_i\sim N(0,{\sigma }^2)& 各{\epsilon }_i相互独立\end{array}\}.$$
• 考虑$a,b$的函数$$Q(a,b)=\sum _{i=1}^n(y_i-a-b{x}_i{)}^2.$$
• 用最小二乘法估计$a,b$，使$Q(a,b)$最小，即$$\hat{a},\hat{b}=\underset{a,b}{\mathrm{argmin}}Q(a,b).$$
• 求导，令偏导数为零，解方程组$$\{\begin{array}{l}\frac{\partial Q}{\partial a}=-2\sum _{i=1}^n(y_i-a-b{x}_i)=0,\\ \frac{\partial Q}{\partial b}=-2\sum _{i=1}^n(y_i-a-b{x}_i)x_i=0.\end{array}$$
• 解出$a,b$，即$\hat{a},\hat{b}$ $$\{\begin{array}{l}\hat{b}=S_{xy}/S_{xx},\\ \hat{a}=\overline{y}-\overline{x}\hat{b},\end{array}$$其中，$$\{\begin{array}{l}{S}_{xy}=\sum _{i=1}^n{x}_i{y}_i-n\overline{x}\overline{y}=\sum _{i=1}^n(x_i-\overline{x})(y_i-\overline{y}),\\ S_{xx}=\sum _{i=1}^n{x}_i^2-n{\overline{x}}^2=\sum _{i=1}^n(x_i-\overline{x}{)}^2.\end{array}$$
• 所求线性回归方程为$$\hat{y}=\hat{a}+\hat{b}x.$$

2.线性假设的显著性分析(方差分析，F检验法)

1. 给出$$\{\begin{array}{lcr}原假设& H_0:& b=0,\\ 备择假设& H_1:& b\ne 0.\end{array}$$
2. 对原假设$H_0:b=0$，备择假设$H_1:b\ne 0$，选统计量$$F=\frac{S_{regression}}{S_{error}/n-2}\sim F(1,n-2).$$
3. 对于给出的$\alpha$，查表得$F_{\alpha }(1,n-2)$；由样本值计算出$S_{regression}$和$S_{error}$，从而计算出$F$值。
   方差分析表

方差来源	平方和	自由度	均方	F比
回归	$S_{regression}$	$$1$$	$$M{S}_{regression}=\frac{S_{regression}}{1}$$	$$F=\frac{S_{regression}}{S_{error}/n-2}$$
残差	$$S_{error}$$	$$n-2$$	$$M{S}_{error}=\frac{S_{error}}{n-2}$$
总和	$$S_{yy}$$	$$n-1$$

4. 判断
   （1）若$F>F_{\alpha }(1,n-2)$，则拒绝$H_0$，接受$H_1$，说明回归效果显著；
   （2）若$F<F_{\alpha }(1,n-2)$，则接受$H_0$，说明回归效果不显著。

记号说明
$$\begin{array}{rl}{S}_{yy}& =\sum _{i=1}^n(y_i-\overline{y}{)}^2=\sum _{i=1}^n{y}_i^2-n{\overline{y}}^2\\ & =\sum _{i=1}^n(\widehat{y_i}-\overline{y}{)}^2+\sum _{i=1}^n(y_i-\widehat{y_i}{)}^2+2\sum _{i=1}^n(\widehat{y_i}-\overline{y})(y_i-\widehat{y_i})\\ & =S_{regression}+S_{error}+0,\\ S_{regression}& =(\hat{b}{)}^2{S}_{xx},S_{error}=S_{yy}-S_{regression}=S_{yy}-(\hat{b}{)}^2{S}_{xx},\end{array}$$

$$\begin{gathered} F=\frac{S_{regression}}{S_{error}/n-2}=\frac{\frac{S_{regression}}{{\sigma }^2}/1}{\frac{S_{error}}{{\sigma }^2}/n-2}\sim F(1,n-2), \\ \frac{S_{regression}}{{\sigma }^2}\sim {\chi }^2(1)\frac{S_{error}}{{\sigma }^2}\sim {\chi }^2(n-2). \end{gathered}$$
注:线性假设的显著性分析还可用$T$检验法。
二者($T$检验法和$F$检验法)区别：

• $T$检验法
  对各个回归系数的检验，可以看出哪些回归系数对回归效果更显著，从而忽略那些对因变量贡献小的回归系数。
• $F$检验法
  对所有回归系数的检验，代表进行回归的所有自变量的回归系数的一个总体检验。

PS：用EXCEL进行线性回归分析（数据来源：应用概率统计 陈魁 清华大学出版社 第11章 回归分析 例11.1.2）