平差之经验离散度测量

离散度测量说白了就是方差

理论的离散度测量:

$$随机误差=观测值-期望值$$
$${\epsilon }_j=\underset{\text{Actual }}{\underbrace{l_j}}-\underset{\text{Nominal }}{\underbrace{{\mu }_l}}\Rightarrow \underset{\text{for all }j}{\underbrace{{\forall }_j}}$$
$${\sigma }_l^2=E\left({\epsilon }^2\right)=\underset{n\to \infty }{\lim }\left\{\frac{1}{n}\sum _{j=1}^n{\epsilon }_j^2\right\}=\underset{n\to \infty }{\lim }\left\{\frac{1}{n}{\epsilon }_{1,n}^T\cdot {\epsilon }_{n,1}\right\}$$
$${\sigma }_x^2={\int }_{-\infty }^{\infty }{\left(x-{\mu }_x\right)}^{2}f(x)\mathrm{d}x$$

经验的离散度度测量

已知期望值${\mu }_l$

$$l_{1,n}={\left[\begin{array}{llll}{l}_1& l_2& \cdots & l_n\end{array}\right]}^T\text{ with }n\ll \infty$$
$${\epsilon }_{n,1}=l_{n,1}-e_{n,1}\cdot {\mu }_l$$
$$\begin{array}{c}{s}_l^2={\epsilon }_1^2+{\epsilon }_2^2+{\epsilon }_3^2+\cdots +{\epsilon }_n^2=\frac{1}{n}\cdot {\sum }_{j=1}^n{\epsilon }_j^2\\ =\frac{1}{n}\cdot {\epsilon }_{1,n}^T\cdot {\epsilon }_{n,1}\end{array}$$
单独观测的经验方差:
$$s_l=\sqrt{s_l^2}$$
$$E\left(s_l^2\right)={\sigma }_l^2$$
但是
$$E\left(s_l\right)\ne {\sigma }_l$$
通常
$$E\left(s_l\right)<{\sigma }_l$$

未知期望值${\mu }_l$

$$L_{1,n}={\left[\begin{array}{llll}{l}_1& l_2& \cdots & l_n\end{array}\right]}^T\text{ with }n\ll \infty$$
期望值是不知道的:
只能用样本均值替代总体期望值:
$$\bar{l}=\frac{1}{n}\cdot \sum _{j=1}^n{l}_j=\frac{1}{n}\cdot e_{1,n}^T\cdot L_{n,1}$$
这里引入新的变量$v_j$,表示改正数:
$$改正数=均值-观测值$$
$$v_j=\bar{l}-l_j\dots {\forall }_j\dots \text{ for all }j$$
需满足,因为随机误差的和等于0,这里改正数其实也相当于随机误差了:
$$\sum _{j=1}^n{v}_j=0$$
改正数的向量表达:
$$v_{n,1}=\left[\begin{array}{c}{v}_1\\ v_2\\ v_3\\ ⋮\\ v_n\end{array}\right]=\left[\begin{array}{c}\bar{l}-l_1\\ \bar{l}-l_2\\ \bar{l}-l_3\\ ⋮\\ \bar{l}-l_n\end{array}\right]=e_{n,1}\cdot \bar{l}-L_{n,1}$$
方差(注意是n-1,少一个自由度,因为这里的v是用$\bar{l}$和L算出来的,而$\bar{l}$又是用L算出来的):
$$\begin{array}{rl}{s}_l^2& =\left(v_1^2+v_2^2+v_3^2+\cdots +v_n^2\right)\\ & =\frac{1}{n-1}\cdot \sum _{j=1}^n{v}_i^2\\ & =\frac{1}{n-1}\cdot v_{1,n}^T\cdot v_{1,n}\end{array}$$
$$\epsilon \left(s_l^2\right)={\sigma }_l^2$$
$$E\left(s_i\right)\ne {\sigma }_l$$
$$E\left(s_i\right)<{\sigma }_l$$