概率统计D 07.03 正态总体均值与方差的置信区间

$\color{blue}{\S 7.3 正态总体均值与方差的置信区间}$

$\color{blue}{一、单个正态总体均值的置信区间}$

$设(X_1, X_2, \cdots, X_n)是来自正态总体N(\mu, \sigma^2)的样本.$
$1.设\sigma^2已知,求\mu的置信度为1 - \alpha的置信度区间.$
$\left( \overline X - \dfrac{\sigma}{\sqrt{n}} \mu _{\frac{\alpha}{2}}, \overline X + \dfrac{\sigma}{\sqrt{n}} \mu _{\frac{\alpha}{2}} \right)$

$2.设\sigma^2未知,求\mu的置信度为1 - \alpha的置信区间.$
$由于S^2是\sigma^2的无偏估计,因此用S^2代替\sigma^2,有\\ t = \dfrac{\overline X - \mu}{\dfrac{S}{\sqrt{n}}} \sim t(n-1)$
$由附表3查得t_{\frac{\alpha}{2}}(n-1),有$
$P \left\lbrace \left|\dfrac{\overline X - \mu}{\dfrac{S}{\sqrt{n}} } \right|< t_{\frac{\alpha}{2}}(n-1) \right\rbrace = 1 - \alpha,$
$即P \left \lbrace \overline X - \dfrac{S}{\sqrt{n}}t_{\frac{\alpha}{2}}(n-1) < \mu < \overline X + \dfrac{S}{\sqrt{n}}t_{\frac{\alpha}{2}}(n-1) \right \rbrace = 1 - \alpha ,$
$于是得到\mu的1 - \alpha 置信区间为$
$\left ( \overline X - \dfrac{S}{\sqrt{n}}t_{\frac{\alpha}{2}}(n-1) , \overline X + \dfrac{S}{\sqrt{n}}t_{\frac{\alpha}{2}}(n-1) \right )$

$例1.某车间生产的螺杆直径服从正态分布N(\mu, \sigma^2),今随机\\ 地从中抽取5只测得直径值为22.3, 21.5, 22.0, 21.8, 21.4.$
$(1) 已知\sigma = 0.3, 求\mu 的0.95置信区间;$
$(2) 如果\sigma 未知,求\mu的0.95置信区间.$
$解:n = 5, \overline x = \dfrac{1}{5} \sum \limits _ {i=1}^5 x_i = 21.8.$
$(1) 已知\sigma = 0.3, \alpha = 0.05,查表得\mu_{\frac{\alpha}{2}} = 1.96,\\ 因此\mu 的0.95置信区间为$
$(\overline x - \dfrac{\sigma}{\sqrt{n}} \mu_{\frac{\alpha}{2}}, \overline x + \dfrac{\sigma}{\sqrt{n}} \mu_{\frac{\alpha}{2}}) \\ = (21.8 - \dfrac{0.3}{\sqrt{5}} \times 1.96, 21.8 + \dfrac{0.3}{\sqrt{5}} \times 1.96) \\ = (21.537, 22.063)$
$(2) \sigma未知,s^2 = \dfrac{1}{5-1} \sum \limits _ {i=1}^5 (x_i - \overline x)^2 = 0.135, \\ s = 0.367.查表得t_{\frac{\alpha}{2}}(5-1) = t_{0.025}(4) = 2.7764,\\ 因此\mu的0.95置信区间为\\ (\overline x - \dfrac{s}{\sqrt{n}}t_{\frac{\alpha}{2}}(n-1), \overline x + \dfrac{s}{\sqrt{n}}t_{\frac{\alpha}{2}}(n-1)) \\ = (21.8 - \dfrac{0.367}{\sqrt{5}} \times 2.7764 , 21.8 + \dfrac{0.367}{\sqrt{5}} \times 2.7764) \\ = (21.345, 22.255)$

$\color{blue}{二、单个正态总体方差的置信区间}$

$1.设\mu已知,求\sigma^2的置信度为1 - \alpha的置信区间$
$由于$
$\chi^2 = \dfrac{1}{\sigma^2} \sum \limits _ {i=1}^n(X_i - \mu)^2 \sim \chi^2(n)$
$对于给定置信度1 - \alpha,查表可得\chi^2_{1 - \frac{\alpha}{2}}(n)及\chi^2_{\frac{\alpha}{2}}(n),使$
$P \left \lbrace \chi^2_{1 - \frac{\alpha}{2}}(n) < \dfrac{1}{\sigma^2} \sum \limits _ {i=1}^n (X_i - \mu)^2 < \chi^2_{\frac{\alpha}{2}}(n) \right \rbrace = 1 - \alpha$
$P \left \lbrace \dfrac{\sum \limits _{i=1}^n(X_i - \mu)^2}{\chi^2_{\frac{\alpha}{2}}(n)} < \sigma^2 < \dfrac{\sum \limits _{i=1}^n (X_i - \mu)^2}{\chi^2_{1-\frac{\alpha}{2}}(n)} \right \rbrace = 1 - \alpha,$
$因此,\sigma^2的1-\alpha的置信区间为$
$\left ( \dfrac{\sum \limits _{i=1}^n(X_i - \mu)^2}{\chi^2_{\frac{\alpha}{2}}(n)} , \dfrac{\sum \limits _{i=1}^n (X_i - \mu)^2}{\chi^2_{1-\frac{\alpha}{2}}(n)} \right )$

$2.设\mu未知,求\sigma^2的置信度为1 - \alpha的置信区间$
$由于$
$\dfrac{(n-1)S^2}{\sigma^2} \sim \chi^2(n-1)$
$对于给定置信度1 - \alpha,查表可得\chi^2_{1 -\frac{\alpha}{2}}(n-1)及\chi^2_{\frac{\alpha}{2}}(n-1), 使得\\\\$
$P \left \lbrace \chi^2_{1-\frac{\alpha}{2}}(n-1) < \dfrac{(n-1)S^2}{\sigma^2} < \chi^2_{\frac{\alpha}{2}}(n-1) \right \rbrace = 1 - \alpha$
$即P\left \lbrace \dfrac{(n-1)S^2}{\chi^2_{\frac{\alpha}{2}}(n-1)} < \sigma^2 < \dfrac{(n-1)S^2}{\chi^2_{1 - \frac{\alpha}{2}}(n-1)} \right \rbrace = 1 - \alpha$
$因此,方差\sigma^2的置信度为1 - \alpha的置信区间为$
$\left ( \dfrac{(n-1)S^2}{\chi^2_{\frac{\alpha}{2}}(n-1)} , \dfrac{(n-1)S^2}{\chi^2_{1 - \frac{\alpha}{2}}(n-1)} \right )$

$而标准差\sigma的1 - \alpha的置信区间为$
$\left ( \dfrac{\sqrt{n-1}S}{\sqrt{\chi^2_{\frac{\alpha}{2}}(n-1)}}, \dfrac{\sqrt{n-1}S}{\sqrt{\chi^2_{1- \frac{\alpha}{2}}(n-1)}} \right )$

$例2.从正态总体N(\mu, \sigma^2)中抽取容量为5的样本,\\ 其观测值为1.86, 3.22, 1.46, 4.01, 2.64 \\ (1) 已知\mu = 3, 求\sigma^2 的0.95置信区间;\\ (2) 如果\mu未知,求\sigma^2的0.95置信区间.$
$解:n = 5, \alpha = 0.05$
$(1) 已知\mu = 3, 查表得\chi^2_{1 - \frac{\alpha}{2}}(n) = \chi^2_{0.975}(5) = 0.831, \\ \chi^2_{\frac{\alpha}{2}}(n) = \chi^2_{0.025}(5) = 12.833.\\ 由已知数据算得\sum \limits _{i=1}^5(x_i-\mu)^2 = 4.8693,\\ 因此\sigma^2的0.95置信区间为$
$\left ( \dfrac{\sum \limits _{i=1}^n(X_i - \mu)^2}{\chi^2_{\frac{\alpha}{2}}(n)} , \dfrac{\sum \limits _{i=1}^n (X_i - \mu)^2}{\chi^2_{1-\frac{\alpha}{2}}(n)} \right )$
$=\left ( \dfrac{4.8693}{12.833}, \dfrac{4.8693}{0.831}\right ) \\ = (0.379, 5.860)$
$(2) \mu未知,查表得\chi^2_{1 - \frac{\alpha}{2}}(n-1) = \chi^2_{0.975}(4) = 0.484, \\ \chi^2_{\frac{\alpha}{2}}(n-1) = \chi^2_{0.025}(4) = 11.143, 由已知数据算得\overline x = 2.64, \\ (n-1)s^2 = 4.2141, 因此\sigma^2的0.95置信区间为$
$\left ( \dfrac{\sqrt{n-1}S}{\sqrt{\chi^2_{\frac{\alpha}{2}}(n-1)}}, \dfrac{\sqrt{n-1}S}{\sqrt{\chi^2_{1- \frac{\alpha}{2}}(n-1)}} \right )$
$= \left ( \dfrac{4.2141}{11.143}, \dfrac{4.2141}{0.484} \right) \\ = (0.378, 8.707)$

$\color{blue}{三、两个正态总体均值差的置信区间}$

$设总体X \sim N(\mu_1, \sigma_1^2),总体Y \sim N(\mu_2, \sigma_2^2), X与Y独立,\\ (X_1, X_2, \cdots, X_m)与(Y_1, Y_2, \cdots, Y_n)分别来自X与Y的\\ 相互独立的样本,并设它们的样本均值分别为\overline X,\overline Y, 样本方差\\ 分别为S_1^2, S_2^2.$

$1.设\sigma_1^2和\sigma_2^2都已知,求\mu_1 - \mu_2的置信度为1 - \alpha的置信区间.$
$由于\overline X 与 \overline Y相互独立,且\overline X \sim N(\mu_1, \dfrac{\sigma_1^2}{m}), \overline Y \sim N(\mu_2, \dfrac{\sigma_2^2}{n}),\\ 于是可知\overline X - \overline Y \sim N(\mu_1 - \mu_2, \dfrac{\sigma_1^2}{m} + \dfrac{\sigma_2^2}{n}),从而 \\ \mu = \dfrac{\overline X - \overline Y - (\mu_1 - \mu_2)}{\sqrt{\dfrac{\sigma_1^2}{m} + \dfrac{\sigma_2^2}{n}}} \sim N(0, 1)$
$对于给定的置信度1 - \alpha,查表得\mu_{\frac{\alpha}{2}},使$
$P \left \lbrace \dfrac{|\overline X - \overline Y - (\mu_1 - \mu_2)|}{\sqrt{\dfrac{\sigma_1^2}{m} + \dfrac{\sigma_2^2}{n}}} < \mu_{\frac{\alpha}{2}} \right \rbrace = 1 - \alpha$

$即P \left \lbrace \overline X - \overline Y - \sqrt{\dfrac{\sigma_1^2}{m} + \dfrac{\sigma_2^2}{n}} \mu_{\frac{\alpha}{2}} < \mu_1 - \mu_2 < \overline X - \overline Y + \sqrt{\dfrac{\sigma_1^2}{m} + \dfrac{\sigma_2^2}{n}} \mu_{\frac{\alpha}{2}} \right \rbrace = 1 - \alpha$

$从而得到\mu_1 - \mu_2 的置信度为1 - \alpha的置信区间为$

$\left ( \overline X - \overline Y - \sqrt{\dfrac{\sigma_1^2}{m} + \dfrac{\sigma_2^2}{n}} \mu_{\frac{\alpha}{2}} , \overline X - \overline Y + \sqrt{\dfrac{\sigma_1^2}{m} + \dfrac{\sigma_2^2}{n}} \mu_{\frac{\alpha}{2}} \right )$

$2.设\sigma_1^2 = \sigma_2^2 = \sigma^2为未知,求\mu_1 - \mu_2的置信度为1 - \alpha的置信区间.$
$\left ( \overline X - \overline Y -t_{\frac{\alpha}{2}}(m+n-2) S_w\sqrt{\dfrac{1}{m} + \dfrac{1}{n}}, \overline X - \overline Y +t_{\frac{\alpha}{2}}(m+n-2) S_w\sqrt{\dfrac{1}{m} + \dfrac{1}{n}} \right)$
$其中S_w^2 = \dfrac{(m-1)S_1^2 + (n-1)S_2^2}{m+n-2}$

$例3.设总体X \sim N(\mu_1, 4), 总体Y \sim N(\mu_2, 6),分别独立地从\\ 这两个总体中抽取样本,样本容量依次为16和24,样本均值依次为\\ 16.9和15.3,求两个总体均值差\mu_1 - \mu_2的置信度为0.95的置信区间.$
$解:由题设可知m = 16, n = 24, \overline x = 16.9, \overline y = 15.3, \\ \sigma_1^2 = 4, \sigma_2^2 = 6, 1 - \alpha = 0.95, \alpha = 0.05, \\ 查附表1得\mu_{\frac{\alpha}{2}} = \mu_{0.025} = 1.96.从而可得\mu_1 - \mu_2 \\ 的置信度为0.95的置信区间为$
$\left ( 16.9 - 15.3 - 1.96 \times \sqrt{\dfrac{4}{16} + \dfrac{6}{24}}, 16.9 - 15.3 + 1.96 \times \sqrt{\dfrac{4}{16} + \dfrac{6}{24}} \right) \\ = (0.214, 2.986)$

$例4.为了估计磷肥对某种农作物增产的作用,选20块条件大致相同的\\ 地块进行对比试验.其中10块地施磷肥,另外10块地不施磷肥,得到单\\ 位面积的产量(单位:kg)如下:\\ 施磷肥: 620, 570, 650, 600, 630, 580, 570, 600, 600, 580 \\ 不施磷肥:560, 590, 560, 570, 580, 570, 600, 550, 570, 550 \\ 设施磷肥的地块单位面积产量X \sim N(\mu_1, \sigma^2),不施磷肥的地块单位\\ 面积产量Y \sim N(\mu_2, \sigma^2).求\mu_1 - \mu_2的置信度为0.95的置信区间.$
$解:由题设,两个正态总体的方差相等,但\sigma^2未知, m=10, n = 10, \\ \alpha = 0.05, 1 - \alpha = 0.95, \overline x = 600, \overline y = 570, \\ s_1^2 = \dfrac{6400}{9}, s_2^2 = \dfrac{2400}{9}, s_w = \sqrt{\dfrac{(m-1)s_1^2 + (n-1)s_2^2}{m+n-2}} = 22,\\ 查表得t_{\frac{\alpha}{2}}(m+n-2) = t_{0.025}(18) = 2.1009, \\ 因此\mu_1 - \mu_2的置信度为0.95的置信区间为$
$\left ( \overline x - \overline y - s_w\sqrt{\dfrac{1}{m} + \dfrac{1}{n}} t_{\frac{\alpha}{2}}(m+n-2), \overline x - \overline y + s_w \sqrt{\dfrac{1}{m} + \dfrac{1}{n} t_{\frac{\alpha}{2}}(m+n-2)} \right)$
$= \left ( 600 - 570 - 22 \times \sqrt{\dfrac{1}{10} + \dfrac{1}{10}} \times 2.1009, 600 - 570 + 22 \times \sqrt{\dfrac{1}{10} + \dfrac{1}{10}} \times 2.1009 \right)$
$=(9, 51)$

$\color{blue}{四、两个正态总体方差比的置信区间}$

$1.设\mu_1, \mu_2均已知,求\sigma_1^2/\sigma_2^2的置信水平为1 - \alpha的置信区间.$
$由于\dfrac{1}{\sigma_1^2} \sum \limits _ {i=1}^m (X_i - \mu_1)^2 与\dfrac{1}{\sigma_2^2} \sum \limits _{j=1}^n(Y_j - \mu_2)^2相互独立,且$
$\dfrac{1}{\sigma_1^2} \sum \limits _ {i=1}^m (X_i - \mu_1)^2 \sim \chi^2(m),$
$\dfrac{1}{\sigma_2^2} \sum \limits _ {j=1}^n (X_j - \mu_2)^2 \sim \chi^2(n),$
$从而可知$
$F = \dfrac{\sum \limits _ {i=1}^m(X_i - \mu_1)^2}{\sum \limits _ {j=1}^n(Y_j - \mu_2)^2} \cdot \dfrac{n\sigma_2^2}{m\sigma_1^2} \sim F(m,n)$
$对于给定的置信度1 - \alpha,查表得F_{1-\frac{\alpha}{2}}(m,n)及F_{\frac{\alpha}{2}}(m,n),使$
$P \left \lbrace F_{1-\frac{\alpha}{2}}(m,n) < \dfrac{\sum \limits _ {i=1}^m(X_i - \mu_1)^2}{\sum \limits _ {j=1}^n(Y_j - \mu_2)^2} \cdot \dfrac{n\sigma_2^2}{m\sigma_1^2} < F_{\frac{\alpha}{2}}(m,n) \right \rbrace = 1 - \alpha$
$即$
$P \left \lbrace \dfrac{n \sum \limits _ {i=1}^m(X_i - \mu_1)^2}{m \sum \limits _ {j=1}^n(Y_j - \mu_2)^2} \dfrac{1}{F_{\frac{\alpha}{2}}(m,n)} < \dfrac{\sigma_1^2}{\sigma_2^2} < \dfrac{n \sum \limits _ {i=1}^m(X_i - \mu_1)^2}{m \sum \limits _ {j=1}^n(Y_j - \mu_2)^2} \dfrac{1}{F_{1 - \frac{\alpha}{2}}(m,n)} \right \rbrace = 1 - \alpha$
$因此,\dfrac{\sigma_1^2}{\sigma_2^2}的置信度为1 - \alpha的置信区间为$
$\left( \dfrac{n \sum \limits _ {i=1}^m(X_i - \mu_1)^2}{m \sum \limits _ {j=1}^n(Y_j - \mu_2)^2} \dfrac{1}{F_{\frac{\alpha}{2}}(m,n)} , \dfrac{n \sum \limits _ {i=1}^m(X_i - \mu_1)^2}{m \sum \limits _ {j=1}^n(Y_j - \mu_2)^2} \dfrac{1}{F_{1 - \frac{\alpha}{2}}(m,n)} \right)$

$2.设\mu_1与\mu_2都未知,求\sigma_1^2/\sigma_2^2的置信水平为1-\alpha的置信区间$
$\left( \dfrac{S_1^2}{S_2^2} \cdot \dfrac{1}{F_{\frac{\alpha}{2}}(m-1, n-1)}, \dfrac{S_1^2}{S_2^2} F_{\frac{\alpha}{2}}(n-1, m-1) \right)$

$例5.设总体X \sim N(24, \sigma_1^2),总体Y \sim N(20, \sigma_2^2).\\ 从总体X和Y中独立地抽的样本值如下:$
$总体X:23, 22, 26, 24, 22, 25;$
$总体Y:22, 18, 19, 23, 17$
$求\sigma_1^2/\sigma_2^2的置信度为0.95的置信区间.$
$解:已知\mu_1 = 24, m = 6; \mu_2 = 20, n = 5.由已知数据可算得$
$\sum \limits _{i=1}^6(x_i - 24)^2 = 14, \sum \limits _{j=1}^5(y_j - 20)^2 = 27.$
$因1 - \alpha = 0.95,故\alpha = 0.05.查附表5,\\ 可得F_{0.025}(6, 5) = 6.98, F_{0.025}(5, 6) = 5.99$
$从而可得\sigma_1^2/\sigma_2^2 的置信度为0.95的置信区间为$
$\left( \dfrac{5 \times 14}{6 \times 27 \times 6.98}, \dfrac{5 \times 14 \times 5.99}{6 \times 27} \right) = (0.06, 2.59)$

$例6.从参数\mu_1, \mu_2, \sigma_1^2, \sigma_2^2都未知的两正态总体N(\mu_1, \sigma_1^2),N(\mu_2, \sigma_2^2)\\ 中分别独立地抽取样本,它们的样本容量分别为m=10, n=8,样本方\\ 差分别为s_1^2 = 3.6, s_2^2 = 2.8,求二总体方差比\sigma_1^2/\sigma_2^2的置信度为0.95的\\ 置信区间.$
$解:这里1 - \alpha = 0.95, \alpha = 0.05,查F分布表得:\\ F_{\frac{\alpha}{2}}(m-1, n-1) = F_{0.025}(9, 7) = 4.82, \\ F_{\frac{\alpha}{2}}(n-1, m-1) = F_{0.025}(7, 9) = 4.20, \\ \sigma_1^2/\sigma_2^2的置信度为0.95的置信区间为$
$\left( \dfrac{s_1^2}{s_2^2} \cdot \dfrac{1}{F_{\frac{\alpha}{2}}(m-1, n-1)}, \dfrac{s_1^2}{s_2^2} \cdot F_{\frac{\alpha}{2}}(n-1, m-1) \right)$
$=\left(\dfrac{3.6}{2.8} \times \dfrac{1}{4.82}, \dfrac{3.6}{2.8} \times 4.20 \right)$
$= (0.27, 5.42)$