线性测量的参数最小二乘法处理
1、最小二乘原理
两种方法——代数法和矩阵法。最小二乘原理需满足测量误差无偏、正态分布和相对独立三个基本原则
(1) 代数法
设直接量 Y 1 , Y 2 , . . . , Y n Y_1,Y_2,...,Y_n Y1,Y2,...,Yn的估计量分别为 y 1 , y 2 , . . . , y n y_1,y_2,...,y_n y1,y2,...,yn: { y 1 = f 1 ( x 1 , x 2 , . . . , x t ) y 2 = f 2 ( x 1 , x 2 , . . . , x t ) . . . y n = f n ( x 1 , x 2 , . . . , x t ) \left\{ \begin{array}{lr} y_1=f_1(x_1,x_2,...,x_t)& \\ y_2=f_2(x_1,x_2,...,x_t)\\ ...\\ y_n=f_n(x_1,x_2,...,x_t)& \end{array} \right. ⎩ ⎨ ⎧y1=f1(x1,x2,...,xt)y2=f2(x1,x2,...,xt)...yn=fn(x1,x2,...,xt)测量数据 l 1 , l 2 , . . . , l n l_1,l_2,...,l_n l1,l2,...,ln的残余误差为 { v 1 = l 1 − y 1 = l 1 − f 1 ( x 1 , x 2 , . . . , x t ) v 2 = l 2 − y 2 = l 2 − f 2 ( x 1 , x 2 , . . . , x t ) . . . v n = l n − y n = l n − f n ( x 1 , x 2 , . . . , x t ) \left\{ \begin{array}{lr} v_1=l_1-y_1=l_1-f_1(x_1,x_2,...,x_t)& \\ v_2=l_2-y_2=l_2-f_2(x_1,x_2,...,x_t)\\ ...\\ v_n=l_n-y_n=l_n-f_n(x_1,x_2,...,x_t)& \end{array} \right. ⎩ ⎨ ⎧v1=l1−y1=l1−f1(x1,x2,...,xt)v2=l2−y2=l2−f2(x1,x2,...,xt)...vn=ln−yn=ln−fn(x1,x2,...,xt)使 v 1 2 σ 1 2 + v 2 2 σ 2 2 + . . . + v n 2 σ n 2 = 最小(核心思想) \frac{v_1^2}{\sigma_1^2}+\frac{v_2^2}{\sigma_2^2}+...+\frac{v_n^2}{\sigma_n^2}=最小(核心思想) σ12v12+σ22v22+...+σn2vn2=最小(核心思想)引入权重( P 1 : P 2 : . . . : P n = 1 σ 1 2 : 1 σ 2 2 : . . . : 1 σ n 2 P_1:P_2:...:P_n=\frac{1}{\sigma_1^2}:\frac{1}{\sigma_2^2}:...:\frac{1}{\sigma_n^2} P1:P2:...:Pn=σ121:σ221:...:σn21)可得: P 1 v 1 2 + P 2 v 2 2 + . . . + P n v n 2 = ∑ i = 1 n P i v i 2 = 最小 P_1v_1^2+P_2v_2^2+...+P_nv_n^2=\sum_{i=1}^{n}P_iv_i^2=最小 P1v12+P2v22+...+Pnvn2=i=1∑nPivi2=最小等精度测量中: v 1 2 + v 2 2 + . . . + v n 2 = ∑ i = 1 n v i 2 = 最小 v_1^2+v_2^2+...+v_n^2=\sum_{i=1}^{n}v_i^2=最小 v12+v22+...+vn2=i=1∑nvi2=最小
(2) 矩阵法
L
=
(
l
1
l
2
.
.
.
l
n
)
L=\left( \begin{matrix} l_1 \\ l_2 \\ ... \\ l_n \\ \end{matrix} \right)
L=
l1l2...ln
x
^
=
(
x
1
x
2
.
.
.
x
t
)
\hat{x}=\left( \begin{matrix} x_1 \\ x_2 \\ ... \\ x_t \\ \end{matrix} \right)
x^=
x1x2...xt
V
=
(
v
1
v
2
.
.
.
v
n
)
V=\left( \begin{matrix} v_1 \\ v_2 \\ ... \\ v_n \\ \end{matrix} \right)
V=
v1v2...vn
A
=
(
a
11
a
12
.
.
.
a
1
t
a
21
a
22
.
.
.
a
2
t
⋮
a
n
1
a
n
2
.
.
.
a
n
t
)
A=\left( \begin{matrix} a_{11} & a_{12} & ... & a_{1t} \\ a_{21} & a_{22} & ... & a_{2t} \\ & &\vdots \\ a_{n1} & a_{n2} & ... & a_{nt}\\ \end{matrix} \right)
A=
a11a21an1a12a22an2......⋮...a1ta2tant
其中,
L
L
L为
n
n
n个直接测量结果(测量数据);
x
^
\hat{x}
x^为
t
t
t个代求的被测量的估计值;
V
V
V为
n
n
n个直接测量结果的残差;
A
A
A为
n
n
n个误差方程的
n
×
t
n\times t
n×t个系数。
误差方差:
V
=
L
−
A
x
^
V=L-A\hat{x}
V=L−Ax^a.等精度测量中
要使
V
T
V
=
最小
V^TV=最小
VTV=最小或
(
L
−
A
x
^
)
T
(
L
−
A
x
^
)
=
最小
(L-A\hat{x})^T(L-A\hat{x})=最小
(L−Ax^)T(L−Ax^)=最小b.不等精度测量中
要使
V
T
P
V
=
最小
V^TPV=最小
VTPV=最小或
(
L
−
A
x
^
)
T
P
(
L
−
A
x
^
)
=
最小
(L-A\hat{x})^TP(L-A\hat{x})=最小
(L−Ax^)TP(L−Ax^)=最小其中
P
P
P为
n
×
n
n\times n
n×n的对角矩阵
2、正规方程
待求参数最小二乘法处理的基本环节
(1) 等精度测量
A T V = 0 A^TV=0 ATV=0有因为 V = L − x ^ V=L-\hat{x} V=L−x^ A T L − A T A x ^ = 0 A^TL-A^TA\hat{x}=0 ATL−ATAx^=0 ( A T A ) x ^ = A T L (A^TA)\hat{x}=A^TL (ATA)x^=ATL令 c = A T A c=A^TA c=ATA,则正规方程可写为 c x ^ = A T L c\hat{x}=A^TL cx^=ATL则 x ^ = c − 1 A T L \hat{x}=c^{-1}A^TL x^=c−1ATL, x ^ \hat{x} x^的数学期望为: E [ x ^ ] = E ( c − 1 A T L ) = c − 1 A T E ( L ) = c − 1 A T Y = c − 1 A T A x = X E[\hat{x}]=E(c^{-1}A^TL)=c^{-1}A^TE(L)=c^{-1}A^TY=c^{-1}A^TAx=X E[x^]=E(c−1ATL)=c−1ATE(L)=c−1ATY=c−1ATAx=X其中 Y = ( Y 1 Y 2 . . . Y n ) , X = ( X 1 X 2 . . . X n ) Y=\left( \begin{matrix} Y_1 \\ Y_2 \\ ... \\ Y_n \\ \end{matrix} \right), X=\left( \begin{matrix} X_1 \\ X_2 \\ ... \\ X_n \\ \end{matrix} \right) Y= Y1Y2...Yn ,X= X1X2...Xn x ^ \hat{x} x^是 X X X的无偏估计
(2) 不等精度测量
A
T
P
V
=
0
A^TPV=0
ATPV=0因为
V
=
L
−
A
x
^
V=L-A\hat{x}
V=L−Ax^,所以
A
T
P
(
l
−
A
x
^
)
=
0
A^TP(l-A\hat{x})=0
ATP(l−Ax^)=0又因为
A
T
P
A
=
A
T
P
L
A^TPA=A^TPL
ATPA=ATPL,所以参数的最小二乘解为:
x
^
=
(
A
T
P
A
)
−
1
A
T
P
L
\hat{x}=(A^TPA)^{-1}A^TPL
x^=(ATPA)−1ATPL令
c
∗
=
A
∗
T
A
∗
=
A
T
P
A
c*=A*^TA*=A^TPA
c∗=A∗TA∗=ATPA
x
^
=
c
∗
−
1
A
T
P
L
\hat{x}=c*^{-1}A^TPL
x^=c∗−1ATPL
x
^
\hat{x}
x^的数学期望:
E
[
x
^
]
=
E
(
c
−
1
A
T
L
)
=
c
−
1
A
T
E
(
L
)
=
c
−
1
A
T
Y
=
c
−
1
A
T
A
x
=
X
E[\hat{x}]=E(c^{-1}A^TL)=c^{-1}A^TE(L)=c^{-1}A^TY=c^{-1}A^TAx=X
E[x^]=E(c−1ATL)=c−1ATE(L)=c−1ATY=c−1ATAx=X正规方程的重要过程:
V
=
L
−
A
x
^
V=L-A\hat{x}
V=L−Ax^
c
∗
=
A
T
P
A
c*=A^TPA
c∗=ATPA
x
^
=
c
∗
−
1
A
T
P
L
\hat{x}=c*^{-1}A^TPL
x^=c∗−1ATPL
3、精度估计
(1) 等精度测量数据的精度估计
σ = ∑ i = 1 n v i 2 n − t \sigma=\sqrt{\frac{\sum_{i=1}^{n}v_i^2}{n-t}} σ=n−t∑i=1nvi2其中 t t t为变量个数, n n n为测量次数
(2) bu 等精度测量数据的精度估计
σ = ∑ i = 1 n P i v i 2 n − t \sigma=\sqrt{\frac{\sum_{i=1}^{n}P_iv_i^2}{n-t}} σ=n−t∑i=1nPivi2
4、最小二乘估计量的精度估计
{ σ x 1 = σ d 11 σ x 2 = σ d 22 . . . σ x t = σ d t t \left\{ \begin{array}{lr} \sigma_{x1}=\sigma \sqrt{d_{11}} \\ \sigma_{x2}=\sigma \sqrt{d_{22}}\\ ...\\ \sigma_{xt}=\sigma \sqrt{d_{tt}} \end{array} \right. ⎩ ⎨ ⎧σx1=σd11σx2=σd22...σxt=σdtt其中 c − 1 = ( A T P A ) − 1 = ( d 11 d 12 . . . d 1 t d 21 d 22 . . . d 2 t ⋮ d t 1 d t 2 . . . d t t ) c^{-1}=(A^TPA)^{-1}=\left( \begin{matrix} d_{11} & d_{12} & ... & d_{1t} \\ d_{21} & d_{22} & ... & d_{2t} \\ & &\vdots \\ d_{t1} & d_{t2} & ... &d_{tt}\\ \end{matrix} \right) c−1=(ATPA)−1= d11d21dt1d12d22dt2......⋮...d1td2tdtt
5、总结(线性测量*)
字母 | 名称 |
---|---|
V | 残差 |
L | 测量数据 |
A | 未知量系数 |
X | 未知量 |
P | 权重 |
V
=
(
v
1
v
2
.
.
.
v
n
)
V=\left( \begin{matrix} v_1 \\ v_2 \\ ... \\ v_n \\ \end{matrix} \right)
V=
v1v2...vn
L
=
(
l
1
l
2
.
.
.
l
n
)
L=\left( \begin{matrix} l_1 \\ l_2 \\ ... \\ l_n \\ \end{matrix} \right)
L=
l1l2...ln
A
=
(
a
11
a
12
.
.
.
a
1
t
a
21
a
22
.
.
.
a
2
t
⋮
a
n
1
a
n
2
.
.
.
a
n
t
)
A=\left( \begin{matrix} a_{11} & a_{12} & ... & a_{1t} \\ a_{21} & a_{22} & ... & a_{2t} \\ & &\vdots \\ a_{n1} & a_{n2} & ... & a_{nt}\\ \end{matrix} \right)
A=
a11a21an1a12a22an2......⋮...a1ta2tant
x
^
=
(
x
1
x
2
.
.
.
x
t
)
\hat{x}=\left( \begin{matrix} x_1 \\ x_2 \\ ... \\ x_t \\ \end{matrix} \right)
x^=
x1x2...xt
P
=
(
P
1
P
2
⋱
P
n
)
P=\left( \begin{matrix} P_1 & & & \\ & P_2 & & \\ & &\ddots\\ & & &P_n \\ \end{matrix} \right)
P=
P1P2⋱Pn
x
^
=
c
−
1
A
T
P
L
\hat{x}=c^{-1}A^TPL
x^=c−1ATPL,其中
c
−
1
=
(
A
T
P
A
)
−
1
c^{-1}=(A^TPA)^{-1}
c−1=(ATPA)−1
V
=
L
−
A
x
^
V=L-A\hat{x}
V=L−Ax^
标准差
σ
=
∑
i
=
1
n
v
i
2
n
−
t
\sigma=\sqrt{\frac{\sum_{i=1}^{n}v_i^2}{n-t}}
σ=n−t∑i=1nvi2,则
x
1
,
x
2
,
.
.
.
,
x
t
x_1,x_2,...,x_t
x1,x2,...,xt的标准差为:
σ
1
=
σ
d
11
,
σ
2
=
σ
d
22
,
.
.
.
,
σ
1
=
σ
d
t
t
\sigma_1=\sigma\sqrt{d_{11}},\sigma_2=\sigma\sqrt{d_{22}},...,\sigma_1=\sigma\sqrt{d_{tt}}
σ1=σd11,σ2=σd22,...,σ1=σdtt其中
c
−
1
=
(
d
11
d
12
.
.
.
d
1
t
d
21
d
22
.
.
.
d
2
t
⋮
d
t
1
d
t
2
.
.
.
d
t
t
)
c^{-1}=\left( \begin{matrix} d_{11} & d_{12} & ... & d_{1t} \\ d_{21} & d_{22} & ... & d_{2t} \\ & &\vdots \\ d_{t1} & d_{t2} & ... &d_{tt}\\ \end{matrix} \right)
c−1=
d11d21dt1d12d22dt2......⋮...d1td2tdtt
例题:
x
−
3
y
=
−
5.6
,
P
1
=
1
;
x-3y=-5.6,P_1=1;
x−3y=−5.6,P1=1;
4
x
+
y
=
8.1
,
P
2
=
2
;
4x+y=8.1, P_2=2;
4x+y=8.1,P2=2;
2
x
−
y
=
0.5
,
P
3
=
3
2x-y=0.5, P_3=3
2x−y=0.5,P3=3解:
L
=
(
−
5.6
8.1
0.5
)
;
A
=
(
1
−
1
4
1
2
−
1
)
;
x
^
=
(
x
y
)
;
P
=
(
1
2
3
)
L=\left( \begin{matrix} -5.6 \\ 8.1 \\ 0.5 \\ \end{matrix} \right);A=\left( \begin{matrix} 1 & -1 \\ 4 &1 \\ 2 & -1 \\ \end{matrix} \right);\hat{x}=\left( \begin{matrix} x \\ y \\ \end{matrix} \right);P=\left( \begin{matrix} 1 & & \\ & 2 & \\ & & 3 \\ \end{matrix} \right)
L=
−5.68.10.5
;A=
142−11−1
;x^=(xy);P=
123
又因为
x
^
=
c
−
1
A
T
P
L
\hat{x}=c^{-1}A^TPL
x^=c−1ATPL
c
−
1
=
(
A
T
P
A
)
−
1
=
(
0.0223
0.0016
0.0016
0.0715
)
c^{-1}=(A^TPA)^{-1}=\left( \begin{matrix} 0.0223 & 0.0016 \\ 0.0016 &0.0715 \\ \end{matrix} \right)
c−1=(ATPA)−1=(0.02230.00160.00160.0715)所以
d
11
=
0.0223
,
d
22
=
0.0715
d_{11}=0.0223,d_{22}=0.0715
d11=0.0223,d22=0.0715。因为
x
^
=
c
−
1
A
T
P
L
=
(
1
,
4345
2.3525
)
\hat{x}=c^{-1}A^TPL=\left( \begin{matrix} 1,4345 \\ 2.3525 \\ \end{matrix} \right)
x^=c−1ATPL=(1,43452.3525)所以
x
=
1.4345
,
y
=
2.352
x=1.4345, y=2.352
x=1.4345,y=2.352。又因为
V
=
L
−
A
x
^
=
(
0.0229
0.0095
−
0.0165
)
V=L-A\hat{x}=\left( \begin{matrix} 0.0229\\ 0.0095\\ -0.0165\\ \end{matrix} \right)
V=L−Ax^=
0.02290.0095−0.0165
所以
σ
=
∑
i
=
1
n
p
i
v
i
2
3
−
2
=
0.039
\sigma=\sqrt{\frac{\sum_{i=1}^{n}p_iv_i^2}{3-2}}=0.039
σ=3−2∑i=1npivi2=0.039
σ
x
=
σ
d
11
=
0.006
,
σ
y
=
σ
d
22
=
0.010
\sigma_x=\sigma\sqrt{d_{11}}=0.006,\sigma_y=\sigma\sqrt{d_{22}}=0.010
σx=σd11=0.006,σy=σd22=0.010
6、非线性测量
函数
y
i
=
f
i
(
x
1
,
x
2
,
.
.
.
,
x
t
)
,
i
=
1
,
2
,
.
.
.
,
n
y_i=f_i(x_1,x_2,...,x_t),i=1,2,...,n
yi=fi(x1,x2,...,xt),i=1,2,...,n,令
{
x
1
=
x
10
+
δ
1
x
2
=
x
20
+
δ
2
.
.
.
x
t
=
x
t
0
+
δ
t
\left\{ \begin{array}{lr} x_1=x_{10}+\delta_1 \\ x_2=x_{20}+\delta_2\\ ...\\ x_t=x_{t0}+\delta_t \end{array} \right.
⎩
⎨
⎧x1=x10+δ1x2=x20+δ2...xt=xt0+δt 其中
δ
i
\delta_i
δi为近似值的偏差
现将函数在
x
10
,
x
20
,
.
.
.
,
x
t
0
x_{10},x_{20},...,x_{t0}
x10,x20,...,xt0处展开,取一次项,则有
f
i
(
x
1
,
x
2
,
.
.
.
,
x
t
)
=
f
i
(
x
10
,
x
20
,
.
.
.
,
x
t
0
)
+
(
∂
f
i
∂
x
1
)
o
δ
1
+
(
∂
f
i
∂
x
2
)
o
δ
2
+
.
.
.
+
(
∂
f
i
∂
x
t
)
o
δ
t
,
(
i
=
1
,
2
,
.
.
.
,
n
)
f_i(x_1,x_2,...,x_t)=f_i(x_{10},x_{20},...,x_{t0})+(\frac{\partial f_i}{\partial x_1})_o\delta_1+(\frac{\partial f_i}{\partial x_2})_o\delta_2+...+(\frac{\partial f_i}{\partial x_t})_o\delta_t,(i=1,2,...,n)
fi(x1,x2,...,xt)=fi(x10,x20,...,xt0)+(∂x1∂fi)oδ1+(∂x2∂fi)oδ2+...+(∂xt∂fi)oδt,(i=1,2,...,n)其中
(
∂
f
i
∂
x
r
)
o
(\frac{\partial f_i}{\partial x_r})_o
(∂xr∂fi)o为函数
f
i
f_i
fi对
x
r
x_r
xr的偏导数在
x
10
,
x
20
,
.
.
.
,
x
t
0
x_{10},x_{20},...,x_{t0}
x10,x20,...,xt0处的取值,令
l
i
′
=
l
i
−
f
i
(
x
10
,
x
20
,
.
.
.
,
x
t
0
)
l'_i=l_i-f_i(x_{10},x_{20},...,x_{t0})
li′=li−fi(x10,x20,...,xt0)
a
i
1
=
(
∂
f
i
∂
x
1
)
o
;
a
i
2
=
(
∂
f
i
∂
x
2
)
o
;
.
.
.
;
a
i
t
=
(
∂
f
i
∂
x
t
)
o
;
a_{i1}=(\frac{\partial f_i}{\partial x_1})_o;a_{i2}=(\frac{\partial f_i}{\partial x_2})_o;...;a_{it}=(\frac{\partial f_i}{\partial x_t})_o;
ai1=(∂x1∂fi)o;ai2=(∂x2∂fi)o;...;ait=(∂xt∂fi)o;则误差方程代成线性方程组:
{
v
1
=
l
1
′
−
(
a
11
δ
1
+
a
12
δ
2
+
.
.
.
+
a
1
t
δ
t
)
v
2
=
l
2
′
−
(
a
21
δ
1
+
a
22
δ
2
+
.
.
.
+
a
2
t
δ
t
)
.
.
.
v
n
=
l
n
′
−
(
a
11
δ
1
+
a
n
2
δ
2
+
.
.
.
+
a
n
t
δ
t
)
\left\{ \begin{array}{lr} v_1=l'_1-(a_{11}\delta_1+a_{12}\delta_2+...+a_{1t}\delta_t) \\ v_2=l'_2-(a_{21}\delta_1+a_{22}\delta_2+...+a_{2t}\delta_t)\\ ...\\ v_n=l'_n-(a_{11}\delta_1+a_{n2}\delta_2+...+a_{nt}\delta_t) \end{array} \right.
⎩
⎨
⎧v1=l1′−(a11δ1+a12δ2+...+a1tδt)v2=l2′−(a21δ1+a22δ2+...+a2tδt)...vn=ln′−(a11δ1+an2δ2+...+antδt)
例题:
c
1
=
0.2071
,
c
2
=
0.2056
,
c
1
+
c
2
=
0.4111
,
c
1
c
2
c
1
+
c
2
=
0.1035
c_1=0.2071,c_2=0.2056,c_1+c_2=0.4111,\frac{c_1c_2}{c_1+c_2}=0.1035
c1=0.2071,c2=0.2056,c1+c2=0.4111,c1+c2c1c2=0.1035
解:
v
1
=
0.2071
−
c
1
v_1=0.2071-c_1
v1=0.2071−c1
v
2
=
0.2056
−
c
2
v_2=0.2056-c_2
v2=0.2056−c2
v
3
=
0.4111
−
(
c
1
+
c
2
)
v_3=0.4111-(c_1+c_2)
v3=0.4111−(c1+c2)
v
4
=
0.1035
−
c
1
c
2
c
1
+
c
2
v_4=0.1035-\frac{c_1c_2}{c_1+c_2}
v4=0.1035−c1+c2c1c2令
v
1
=
v
2
=
0
v_1=v_2=0
v1=v2=0,则
c
10
=
0.2071
,
c
20
=
0.2056
c_{10}=0.2071,c_{20}=0.2056
c10=0.2071,c20=0.2056。则
f
4
(
c
1
,
c
2
)
=
f
4
(
c
10
,
c
20
)
+
∂
f
4
∂
c
1
(
c
1
−
c
10
)
+
∂
f
4
∂
c
2
(
c
2
−
c
20
)
f_4(c_1,c_2)=f_4(c_{10},c_{20})+\frac{\partial f_4}{\partial c_{1}}(c_1-c_{10})+\frac{\partial f_4}{\partial c_{2}}(c_2-c_{20})
f4(c1,c2)=f4(c10,c20)+∂c1∂f4(c1−c10)+∂c2∂f4(c2−c20)
=
0.1034
+
0.2482
(
c
1
−
0.2071
)
+
0.2518
(
c
2
−
0.2056
)
=0.1034+0.2482(c_1-0.2071)+0.2518(c_2-0.2056)
=0.1034+0.2482(c1−0.2071)+0.2518(c2−0.2056)
=
0.2482
c
1
+
0.2518
c
2
+
0.0002277
=0.2482c_1+0.2518c_2+0.0002277
=0.2482c1+0.2518c2+0.0002277此时
L
=
(
0.2071
0.2056
0.4111
0.1035
)
,
A
=
(
1
0
0
1
1
1
0.2482
0.2518
)
L=\left( \begin{matrix} 0.2071 \\ 0.2056 \\ 0.4111 \\ 0.1035\\ \end{matrix} \right),A=\left( \begin{matrix} 1 & 0 \\ 0 & 1 \\ 1 & 1 \\ 0.2482 & 0.2518 \\ \end{matrix} \right)
L=
0.20710.20560.41110.1035
,A=
1010.24820110.2518
c
−
1
=
(
A
T
A
)
−
1
=
(
0.6603
−
0.3400
−
0.3400
0.6597
)
c^{-1}=(A^TA)^{-1}=\left( \begin{matrix} 0.6603 & -0.3400 \\ -0.3400 & 0.6597 \\ \end{matrix} \right)
c−1=(ATA)−1=(0.6603−0.3400−0.34000.6597)所以
d
11
=
0.6603
,
d
22
=
0.6597
d_{11}=0.6603,d_{22}=0.6597
d11=0.6603,d22=0.6597
x
^
=
c
−
1
A
T
L
=
(
0.1904
0.1882
)
\hat{x}=c^{-1}A^TL=\left( \begin{matrix} 0.1904\\ 0.1882\\ \end{matrix} \right)
x^=c−1ATL=(0.19040.1882)可得
c
1
=
0.1904
,
c
2
=
0.1882
c_1=0.1904,c_2=0.1882
c1=0.1904,c2=0.1882。此时
V
=
L
−
A
x
^
,
σ
=
∑
i
=
1
4
v
i
2
4
−
2
V=L-A\hat{x},\sigma=\sqrt{\frac{\sum_{i=1}^{4}v_i^2}{4-2}}
V=L−Ax^,σ=4−2∑i=14vi2
σ
c
1
=
σ
d
11
,
σ
c
2
=
σ
d
22
\sigma_{c1}=\sigma\sqrt{d_{11}},\sigma_{c2}=\sigma\sqrt{d_{22}}
σc1=σd11,σc2=σd22