《应用多元分析》——王学民，第三章（2）

3.10

设 $\mathbf{x}\sim N_2(0,\mathbf{I})$, 其中 $\mathbf{x}={\left(x_1,x_2\right)}^{\prime }$, 试求已知 $x_1+x_2$ 时 $x_1$ 的条件分布。

答案

$$\left(\begin{array}{c}{x}_1\\ x_1+x_2\end{array}\right)=\left(\begin{array}{rr}1& 0\\ 1& 1\end{array}\right)\left(\begin{array}{c}{x}_1\\ x_2\end{array}\right)\sim N_2\left(\left(\begin{array}{r}0\\ 0\end{array}\right),\left(\begin{array}{rr}1& 1\\ 1& 2\end{array}\right)\right)$$

于是
$$\begin{array}{rl}E(x_1\mid x_1+x_2)& =0+1\times \frac{1}{2}\times \left[(x_1+x_2)+0\right]\\ & =\frac{x_1+x_2}{2}\end{array}$$

$$\begin{array}{rl}V(x_1\mid x_1+x_2)& =1-1\times \frac{1}{2}\times 1=\frac{1}{2}\end{array}$$

故已知 $x_1+x_2$ 时 $x_1$ 的条件分布为$N(\frac{x_1+x_2}{2},\frac{1}{2})$。

3.11

试证
$$\sum _{i=1}^n{\left({\mathbf{x}}_i-\bar{\mathbf{x}}\right)}^{\prime }{\mathbf{\Sigma }}^{-1}\left({\mathbf{x}}_i-\bar{\mathbf{x}}\right)=\underset{\mathbf{\mu }}{\min }\sum _{i=1}^n{\left({\mathbf{x}}_i-\mathbf{\mu }\right)}^{\prime }{\Sigma }^{-1}\left({\mathbf{x}}_i-\mathbf{\mu }\right)$$

答案

$$\begin{array}{rl}\sum _{i=1}^n({\mathbf{x}}_i-\mathbf{\mu }{)}^{\prime }({\mathbf{x}}_i-\mathbf{\mu })& =\sum _{i=1}^n{\left[({\mathbf{x}}_i-\bar{\mathbf{x}})+(\bar{\mathbf{x}}\mathbf{\mu })\right]}^{\prime }{\mathbf{\Sigma }}^{-1}\left[({\mathbf{x}}_i-\bar{\mathbf{x}})(\bar{\mathbf{x}}-\mathbf{\mu })\right]\\ & =\sum _{i=1}^n({\mathbf{x}}_i-\bar{\mathbf{x}}{)}^{\prime }{\mathbf{\Sigma }}^{-1}({\mathbf{x}}_i-\bar{\mathbf{x}})+n(\bar{\mathbf{x}}-\mathbf{\mu }{)}^{\prime }{\mathbf{\Sigma }}^{-1}(\bar{\mathbf{x}}-\mathbf{\mu })\end{array}$$

故当$\mathbf{\mu }=\bar{\mathbf{x}}$时，上式达到最小值。

3.12

在习题 $3.2$ 中, $\mathbf{\mu }=(10,4,7{)}^{\prime }$, 试求
(1) $\left(x_1,x_2\right)$ 的边缘分布;
(2) $x_1\mid \left(x_2,x_3\right)$ 和 $\left(x_1,x_2\right)\mid x_3$ 的条件分布;
(3) $x_3$ 给定时, $x_1$ 与 $x_2$ 的偏相关系数;
(4) $x_1$ 与 $\left(x_2,x_3\right)$ 的复相关系数。

答案

(1)

$$\begin{gathered} \left(\begin{array}{c}{x}_1\\ x_2\end{array}\right)=\left(\begin{array}{cc}1& 0\\ 0& 1\end{array}\right)\left(\begin{array}{c}{x}_1\\ x_2\end{array}\right) \\ E(x_1,x_2)=\left(\begin{array}{cc}1& 0\\ 0& 1\end{array}\right)\left(\begin{array}{c}10\\ 4\end{array}\right)=\left(\begin{array}{c}10\\ 4\end{array}\right) \\ V(x_1,x_2)=\left(\begin{array}{cc}1& 0\\ 0& 1\end{array}\right)\left(\begin{array}{cc}9& -3\\ -3& 5\end{array}\right)\left(\begin{array}{cc}1& 0\\ 0& 1\end{array}\right)=\left(\begin{array}{cc}9& -3\\ -3& 5\end{array}\right) \\ {N}_2\left(\left(\begin{array}{c}10\\ 4\end{array}\right),\left(\begin{array}{cc}9& -3\\ -3& 5\end{array}\right)\right) \end{gathered}$$

(2)

由题意得
$$\begin{gathered} \left(\begin{array}{c}{x}_1\\ x_2\\ x_3\end{array}\right)=\left(\begin{array}{ccc}1& 0& 0\\ 0& 1& 0\\ 0& 0& 1\end{array}\right)\left(\begin{array}{c}{x}_1\\ x_2\\ x_3\end{array}\right) \\ E\left(\begin{array}{c}{x}_1\\ x_2\\ x_3\end{array}\right)=\left(\begin{array}{ccc}1& 0& 0\\ 0& 1& 0\\ 0& 0& 1\end{array}\right)\left(\begin{array}{c}10\\ 4\\ 7\end{array}\right)=\left(\begin{array}{c}10\\ 4\\ 7\end{array}\right) \\ V\left(\begin{array}{c}{x}_1\\ x_2\\ x_3\end{array}\right)=\left(\begin{array}{ccc}1& 0& 0\\ 0& 1& 0\\ 0& 0& 1\end{array}\right)\left(\begin{array}{ccc}9& -3& -3\\ -3& 5& 1\\ -& 1& 5\end{array}\right)\left(\begin{array}{ccc}1& 0& 0\\ 0& 1& 0\\ 0& 0& 1\end{array}\right) \end{gathered}$$

$$\begin{array}{rl}E\left(x_1\right|\left(\begin{array}{r}{x}_2\\ x_3\end{array}\right))& =10+(-3,-3){\left(\begin{array}{rr}5& 1\\ 1& 5\end{array}\right)}^{-1}[\left(\begin{array}{r}{x}_2\\ x_3\end{array}\right)-\left(\begin{array}{r}4\\ 7\end{array}\right)]\\ & =-\frac{x_2}{2}-\frac{x_3}{2}+15\frac{1}{2}\end{array}$$

$$\begin{array}{rl}V\left(x_1\right|\left(\begin{array}{r}{x}_2\\ x_3\end{array}\right))& =9-(-3,-3){\left(\begin{array}{rr}5& 1\\ 1& 5\end{array}\right)}^{-1}\left(\begin{array}{r}-3\\ -3\end{array}\right)=6\end{array}$$
故$x_1|\left(\begin{array}{r}{x}_2\\ x_3\end{array}\right)\sim (-\frac{x_2}{2}-\frac{x_3}{2}+15\frac{1}{2},6)$

又因为：
$$\begin{array}{rl}E\left(\left(\begin{array}{r}{x}_1\\ x_2\end{array}\right)\right|x_3)& =\left(\begin{array}{r}10\\ 4\end{array}\right)+\left(\begin{array}{c}-3\\ 1\end{array}\right)\frac{1}{5}(x_3-7)\\ & =\left(\begin{array}{c}-\frac{3}{5}{x}_3+14\frac{1}{5}\\ \frac{1}{5}{x}_3+2\frac{3}{5}\end{array}\right)\end{array}$$

$$\begin{array}{rl}V\left(\left(\begin{array}{r}{x}_1\\ x_2\end{array}\right)\right|x_3)& =\left(\begin{array}{rr}9& -3\\ -3& 5\end{array}\right)-\left(\begin{array}{r}-3\\ 1\end{array}\right)\frac{1}{5}\left(-3,1\right)\\ & =\frac{12}{5}\left(\begin{array}{cc}3& -1\\ -1& 2\end{array}\right)\end{array}$$
故$\left(\begin{array}{r}{x}_1\\ x_2\end{array}\right)|x_3\sim N_2\left(\left(\begin{array}{c}-\frac{3}{5}{x}_3+14\frac{1}{5}\\ \frac{1}{5}{x}_3+2\frac{3}{5}\end{array}\right),\frac{12}{5}\left(\begin{array}{cc}3& -1\\ -1& 2\end{array}\right)\right)$

(3)

由$(2)$知，$x_3$给定时，$\left(\begin{array}{r}{x}_1\\ x_2\end{array}\right)$的偏协方差矩阵为$\frac{12}{5}\left(\begin{array}{rr}3& -1\\ -1& 2\end{array}\right)$，故其偏相关系数为$\frac{-1}{\sqrt{3}\sqrt{2}}=-\frac{\sqrt{6}}{6}$。

(4)

由题意得：

$$\sqrt{\frac{(-3,-3){\left(\begin{array}{rr}5& 1\\ 1& 5\end{array}\right)}^{-1}\left(\begin{array}{r}-3\\ -3\end{array}\right)}{9}}=\frac{\sqrt{3}}{3}$$

3.13

设 $\mathbf{x}\sim N_3(0,\mathbf{\Sigma })$, 其中

$$\mathbf{\Sigma }=\left(\begin{array}{ccc}1& {\rho }_{12}& {\rho }_{13}\\ {\rho }_{12}& 1& {\rho }_{23}\\ {\rho }_{13}& {\rho }_{23}& 1\end{array}\right)$$

试求
(1) $x_3\mid \left(x_1,x_2\right)$ 的条件分布;
(2) 给定 $x_3$ 时, $x_1$ 和 $x_2$ 的偏协方差。

答案

(1)

$$\left(\begin{array}{r}{x}_3\\ x_1\\ x_2\end{array}\right)\sim N_3\left(\left(\begin{array}{r}0\\ 0\\ 0\end{array}\right),\left(\begin{array}{rrr}1& {\rho }_{13}& {\rho }_{23}\\ {\rho }_{13}& 1& {\rho }_{12}\\ {\rho }_{23}& {\rho }_{12}& 1\end{array}\right)\right)$$

$$\begin{array}{rl}E\left(x_3\right|\left(\begin{array}{r}{x}_1\\ x_2\end{array}\right))& =0+({\rho }_{13},{\rho }_{23}){\left(\begin{array}{rr}1& {\rho }_{12}\\ {\rho }_{12}& 1\end{array}\right)}^{-1}[\left(\begin{array}{r}{x}_1\\ x_2\end{array}\right)-\left(\begin{array}{r}0\\ 0\end{array}\right)]\\ & =\frac{{\rho }_{13}-{\rho }_{12}{\rho }_{23}}{1-{\rho }_{12}^2}{x}_1+\frac{{\rho }_{23}-{\rho }_{12}{\rho }_{13}}{1-{\rho }_{12}^2}{x}_2\end{array}$$

$$\begin{array}{rl}V\left(x_3\right|\left(\begin{array}{r}{x}_1\\ x_2\end{array}\right))& =1-({\rho }_{13},{\rho }_{23}){\left(\begin{array}{rr}1& {\rho }_{12}\\ {\rho }_{12}& 1\end{array}\right)}^{-1}\left(\begin{array}{r}{\rho }_{13}\\ {\rho }_{13}\end{array}\right)\\ & =\frac{1-{\rho }_{12}^2-{\rho }_{13}^2-{\rho }_{23}^2+2{\rho }_{12}{\rho }_{13}{\rho }_{23}}{1-{\rho }_{12}^2}\end{array}$$

$$x_3|\left(\begin{array}{r}{x}_1\\ x_2\end{array}\right)\sim N(\frac{{\rho }_{13}-{\rho }_{12}{\rho }_{23}}{1-{\rho }_{12}^2}{x}_1+\frac{{\rho }_{23}-{\rho }_{12}{\rho }_{13}}{1-{\rho }_{12}^2},\frac{1-{\rho }_{12}^2-{\rho }_{13}^2-{\rho }_{23}^2+2{\rho }_{12}{\rho }_{13}{\rho }_{23}}{1-{\rho }_{12}^2})$$

(2)

给定$x_3$时，$\left(\begin{array}{r}{x}_1\\ x_2\end{array}\right)$的偏协方差矩阵为
$$\left(\begin{array}{cc}1& {\rho }_{12}\\ {\rho }_{12}& 1\end{array}\right)-\left(\begin{array}{r}{\rho }_{13}\\ {\rho }_{23}\end{array}\right)({\rho }_{13},{\rho }_{23})=\left(\begin{array}{cc}1-{\rho }_{13}^2& {\rho }_{12}-{\rho }_{13}{\rho }_{13}\\ {\rho }_{12}-{\rho }_{13}{\rho }_{23}& 1-{\rho }_{23}^2\end{array}\right)$$

3.14

设
$$A=\left(\begin{array}{cccc}\frac{1}{\sqrt{4}}& \frac{1}{\sqrt{4}}& \frac{1}{\sqrt{4}}& \frac{1}{\sqrt{4}}\\ \frac{1}{\sqrt{2}}& -\frac{1}{\sqrt{2}}& 0& 0\\ \frac{1}{\sqrt{6}}& \frac{1}{\sqrt{6}}& -\frac{2}{\sqrt{6}}& 0\\ \frac{1}{\sqrt{12}}& \frac{1}{\sqrt{12}}& \frac{1}{\sqrt{12}}& -\frac{3}{\sqrt{12}}\end{array}\right)$$
(1) 试证 ${\mathbf{AA}}^{\prime }=\mathbf{I}$, 即 $\mathbf{A}$ 是一个正交矩阵;
(2) 设 $\mathbf{y}=\mathbf{Ax}$, 其中 $\mathbf{x}\sim N_4\left(\mu 1,{\sigma }^2\mathbf{I}\right),1=(1,1,1,1{)}^{\prime }$, 试证
$$y_2^2+y_3^2+y_4^2=\sum _{i=1}^4{x}_i^2-\frac{1}{4}{\left(\sum _{i=1}^4{x}_i\right)}^2$$
及 $y_1,y_2,y_3,y_4$ 相互独立, 且 $y_1\sim N\left(2\mu ,{\sigma }^2\right),y_i\sim N\left(0,{\sigma }^2\right),i=2,3,4$ 。

答案

（1）$\mathbf{A}$的$4$个行向量均为单位向量且彼此正交，所以$\mathbf{A}$是正交矩阵。

（2）从$\mathbf{y}=\mathbf{A}\mathbf{x}$，可见${\mathbf{y}}^{\prime }\mathbf{y}=(\mathbf{Ax}{)}^{\prime }\mathbf{Ax}={\mathbf{x}}^{\prime }\mathbf{x}$，且$y_1=\frac{1}{2}{\sum }_{i=1}^4{x}_i$，从而
$$\begin{array}{rl}{y}_2^2+y_3^2+y_4^2& ={\mathbf{y}}^{\prime }\mathbf{y}-y_1^2\\ & ={\mathbf{x}}^{\prime }\mathbf{x}-(\frac{1}{2}\sum _{i=1}^4{x}_i{)}^2\\ & =\sum _{i=1}^4{x}_i^2-\frac{1}{4}(\sum _{i=1}^4{x}_i{)}^2\end{array}$$
于是
$$\mathbf{y}=\mathbf{Ax}\sim N_4(\mu \mathbf{A}1,{\sigma }^2\mathbf{A}{\mathbf{A}}^{\prime })=N_4(\mu \mathbf{A}1,{\sigma }^2\mathbf{I})$$

故$y_1,y_2,y_3,y_4$相互独立，且$y\sim N（2\mu ,{\sigma }^2),y\sim N(0,{\sigma }^2),i=2,3,4$。

3.15

设 $\mathbf{x}\sim N_n\left(\mathbf{\mu },{\sigma }^2\mathbf{I}\right),\bar{x}=\frac{1}{n}{\sum }_{i=1}^n{x}_i,(n-1)s^2={\sum }_{i=1}^n{\left(x_i-\bar{x}\right)}^2={\sum }_{i=1}^n{x}_i^2-n{\bar{x}}^2={\sum }_{i=1}^n{x}_i^2-$ $\frac{1}{n}{\left({\sum }_{i=1}^n{x}_i\right)}^2$, 试证 $\bar{x}$ 和 $s^2$ 相互独立。

[ 提示: 证法一。参考习题3.14，令
$$\mathbf{A}=\left(\begin{array}{cccccc}\frac{1}{\sqrt{n}}& \frac{1}{\sqrt{n}}& \frac{1}{\sqrt{n}}& \cdots & \frac{1}{\sqrt{n}}& \frac{1}{\sqrt{n}}\\ \frac{1}{\sqrt{2\cdot 1}}& \frac{-1}{\sqrt{2\cdot 1}}& 0& \cdots & 0& 0\\ \frac{1}{\sqrt{3\cdot 2}}& \frac{1}{\sqrt{3\cdot 2}}-2\frac{-2}{\sqrt{3\cdot 2}}& \cdots & 0& 0& \\ ⋮& ⋮& ⋮& & ⋮& ⋮\\ \frac{1}{\sqrt{n(n-1)}}& \frac{1}{\sqrt{n(n-1)}}& \frac{1}{\sqrt{n(n-1)}}& \cdots & \frac{1}{\sqrt{n(n-1)}}& \frac{-(n-1)}{\sqrt{n(n-1)}}\end{array}\right)$$

• 证法二。利用 $\S 3.2$ 中的性质 $(9)$, 证明 $\bar{x}=\frac{1}{n}{1}^{\prime }\mathbf{x}$ 与 $\left(\begin{array}{c}{x}_1-\bar{x}\\ x_2-\bar{x}\\ ⋮\\ x_n-\bar{x}\end{array}\right)=\left(\mathbf{I}-\frac{1}{n}{11}^{\prime }\right)\mathbf{x}$ 独立, 其中 $1=(1,1,\cdots ,1{)}^{\prime }$]

答案

证法一

易见，$\mathbf{A}$的所有行都是单位向量且彼此正交，故$\mathbf{A}$是正交矩阵。令$\mathbf{y}=\mathbf{Ax}$，于是$V(y)=\mathbf{A}({\sigma }^2\mathbf{I}){\mathbf{A}}^{\prime }={\sigma }^2\mathbf{A}{\mathbf{A}}^{\prime }={\sigma }^2\mathbf{I}$，从而$y\sim N_n(\ast ,{\sigma }^2\mathbf{I})$，故$y_1,y_2,\cdots ,y_n$相互独立。因为$y_1=(\frac{1}{\sqrt{n}}{1}^{\prime })x=\frac{1}{\sqrt{n}}{\sum }_{i=1}^n{x}_i$，且${\sum }_{i=1}^n{y}_i^2=\mathbf{y}{\mathbf{y}}^{\prime }=(\mathbf{Ax}{)}^{\prime }(\mathbf{Ax})={\mathbf{x}}^{\prime }\mathbf{x}={\sum }_{i=1}^n{x}_i^2$，于是
$$\begin{gathered} \bar{x}=\frac{1}{n}\sum _{i=1}^n{x}_i=\frac{1}{\sqrt{n}}{y}_1 \\ (n-1)s^2=\sum _{i=1}^n{x}_i^2-\frac{1}{n}(\sum _{i=1}^n{x}_i{)}^2=\sum _{i=1}^n{y}_i^2-y_1^2=\sum _{i=2}^n{y}_i^2 \end{gathered}$$
所以$\bar{x}$和$s_2$独立。

证法二

因$\mathbf{x}$ 为多元正态变量，且
$$(\frac{1}{n}{1}^{\prime })({\sigma }^2\mathbf{I})(\mathbf{I}-\frac{1}{n}11^{\prime }{)}^{\prime }=\frac{{\sigma }^2}{n}(1^{\prime }-\frac{1}{n}{1}^{\prime }11^{\prime })=0^{\prime }$$
于是由$\S 3.2$中的结论中的性质$(9)$知，$\bar{x}=\frac{1}{n}{1}^{\prime }x$独立，而$s^2$是$\left(\begin{array}{c}{x}_1-\bar{x}\\ x_2-\bar{x}\\ ⋮\\ x_n-\bar{x}\end{array}\right)$的函数，从而$\bar{x}$与$s^2$独立。

3.16(有用结论)

设 $x_1,x_2,\cdots ,x_n$ 是来自 $p$ 维总体 $x$ 的一个样本, 样本协方差矩阵 $\mathbf{S}>0$, 试证 $n>p$ 。

答案

令$\mathbf{B}=(x_1-\bar{x},\cdots ,x_n-\bar{x})$，则
$$\mathbf{S}=\frac{1}{n-1}\sum _{i=1}^n(x_i-\bar{x})(x_i-\bar{x}{)}^{\prime }=\frac{1}{n-1}\mathbf{B}{\mathbf{B}}^{\prime }$$

于是由$\S 1.5$的性质$(8)$知
$$rank(\mathbf{B})rank(\mathbf{B}{\mathbf{B}}^{\prime })=rank(\mathbf{S})=p$$
由于${\sum }_{i=1}^n(x_i-\bar{x})=0$，从而$\mathbf{B}$不是列满秩的，即有$rank(\mathbf{B})<n$，故$n>p$。

3.17

试证 (3.4.1) 式中的最后一个等式。

答案

令$\mathbf{D}=diag(\sqrt{{\sigma }_{11}},\sqrt{{\sigma }_{22}},\cdots ,\sqrt{{\sigma }_{pp}})$，其中${\sigma }_{11},{\sigma }_{22},\cdots ,{\sigma }_{pp}$是${\mathbf{\Sigma }}_{xx}$是$p$个对角线元素，则
$$\begin{gathered} {\mathbf{\sigma }}_{xy}=Cov(x,y)=\mathbf{D}\rho (x,y)\sqrt{{\sigma }_{yy}}=\sqrt{{\sigma }_{yy}}{\mathbf{D\rho }}_{xy} \\ {\mathbf{\Sigma }}_{\mathbf{xx}}=\mathbf{D}{\mathbf{R}}_{\mathbf{xx}}\mathbf{D} \end{gathered}$$
从而
$$\frac{{\mathbf{\sigma }}_{\mathbf{xy}}^{\prime }{\mathbf{\Sigma }}_{\mathbf{xx}}^{-1}{\mathbf{\sigma }}_{\mathbf{xy}}}{{\sigma }_{yy}}=\frac{(\sqrt{{\sigma }_{yy}}{\mathbf{D\rho }}_{xy}{)}^{\prime }({\mathbf{DR}}_{xy}\mathbf{D}{)}^{-1}(\sqrt{{\sigma }_{yy}}{\mathbf{D\rho }}_{xy})}{{\sigma }_{yy}}={{\mathbf{\rho }}^{\prime }}_{xy}{\mathbf{R}}_{xx}^{-1}{\mathbf{\rho }}_{xy}$$

*3.18

试证 (3.4.5) 式。

答案

设$g(\mathbf{x})=a+{\mathbf{b}}^{\prime }(\mathbf{x}-{\mathbf{\mu }}_{\mathbf{x}})$，于是
$$\begin{array}{rl}E[y-g(\mathbf{x}){]}^2& =E\{(y-\bar{y}+[\bar{y}-g(\mathbf{x})]{\}}^2\\ & =E(y-\bar{y}{)}^2+E[\bar{y}-g(\mathbf{x}){]}^2+2E(y-\bar{y})[\bar{y}-g(\mathbf{x})]\end{array}$$
又因为
$$\begin{array}{rl}E(y-\bar{y})[\bar{y}-g(\mathbf{x})]& =E[y-{\mu }_y-{{\mathbf{\sigma }}^{\prime }}_{xy}{\mathbf{\Sigma }}_{xx}^{-1}(\mathbf{x}-{\mathbf{\mu }}_{\mathbf{x}})][{\mu }_y+{\mathbf{\sigma }}_{xy}^{\prime }{\mathbf{\Sigma }}_{xx}^{-1}(\mathbf{x}-{\mathbf{\mu }}_{\mathbf{x}})-a-{\mathbf{b}}^{\prime }(\mathbf{x}-{\mathbf{\mu }}_{\mathbf{x}})]\\ & =E[(y-{\mu }_y)-{\mathbf{\sigma }}_{xy}^{\prime }{\mathbf{\Sigma }}_{xx}^{-1}(\mathbf{x}-{\mathbf{\mu }}_{\mathbf{x}})][({\mathbf{\sigma }}_{xy}^{\prime }{\mathbf{\Sigma }}_{xx}^{-1}-{\mathbf{b}}^{\prime })(\mathbf{x}-{\mathbf{\mu }}_{\mathbf{x}})]\\ & =E[(y-{\mu }_y)-{\mathbf{\sigma }}_{xy}^{\prime }{\mathbf{\Sigma }}_{xx}^{-1}(\mathbf{x}-{\mathbf{\mu }}_{\mathbf{x}})](\mathbf{x}-{\mathbf{\mu }}_{\mathbf{x}}{)}^{\prime }({\mathbf{\Sigma }}_{xx}^{-1}{\mathbf{\sigma }}_{xy}-\mathbf{b})\\ & =[Cov(y,x)-{\mathbf{\sigma }}_{xy}^{\prime }{\mathbf{\Sigma }}_{xx}^{-1}V(x)]({\mathbf{\Sigma }}_{xx}^{-1}{\mathbf{\sigma }}_{xy}-\mathbf{b})\\ & =({\mathbf{\sigma }}_{xy}^{\prime }-{\mathbf{\sigma }}_{xy}^{\prime }{\mathbf{\Sigma }}_{xx}^{-1}{\mathbf{\Sigma }}_{xx})({\mathbf{\Sigma }}_{xx}^{-1}{\mathbf{\sigma }}_{xy}-\mathbf{b})\\ & =0\end{array}$$
故当$g(x)=\bar{x}$时，$E[y-g(x){]}^2$达到最小值$E(y-\bar{y}{)}^2$。