多维情况下均值方差递推公式的推导

多维情况下均值方差递推公式的推导

均值

$$\begin{array}{rl}{\hat{\mu }}_{n+1}& =\frac{1}{n+1}\sum _{n=1}^{n+1}{x}_n\\ & =\frac{1}{n+1}\left(\sum _{n=1}^n{x}_n+x_{n+1}\right)\\ & =\frac{1}{n+1}\sum _{n=1}^n{x}_n+\frac{x_{n+1}}{n+1}\\ & =\frac{n}{n+1}\frac{1}{n}\sum _{n=1}^n{x}_n+\frac{1}{n+1}{x}_{n+1}\\ & =\frac{n}{n+1}{\hat{\mu }}_n+\frac{1}{n+1}{x}_{n+1}\\ & ={\hat{\mu }}_n+\frac{1}{n+1}\left(x_{n+1}-{\hat{\mu }}_n\right)\end{array}$$

方差

$$\begin{array}{rl}{C}_{n+1}& =\frac{1}{n}\sum _{i=1}^{n+1}\left(x_i-{\hat{\mu }}_{n+1}\right){\left(x_i-{\hat{\mu }}_{n+1}\right)}^t\end{array}$$
即
$$\begin{array}{rl}n{C}_{n+1}& =\sum _{i=1}^{n+1}[\left(x_i-{\hat{\mu }}_n\right)+\left({\hat{\mu }}_n-{\hat{\mu }}_{n+1}\right)][\left(x_i-{\hat{\mu }}_n\right)+\left({\hat{\mu }}_n-{\hat{\mu }}_{n+1}\right){]}^t\\ & =\sum _{i=1}^{n+1}(x_i-{\hat{\mu }}_n)(x_i-{\hat{\mu }}_n{)}^t+\sum _{i=1}^{n+1}({\hat{\mu }}_n-{\hat{\mu }}_{n+1})({\hat{\mu }}_n-{\hat{\mu }}_{n+1}{)}^t\\ & +\sum _{i=1}^{n+1}(x_i-{\hat{\mu }}_n)({\hat{\mu }}_n-{\hat{\mu }}_{n+1}{)}^t+\sum _{i=1}^{n+1}({\hat{\mu }}_n-{\hat{\mu }}_{n+1})(x_i-{\hat{\mu }}_n{)}^t\end{array}$$
其中
$$\begin{array}{rl}\sum _{i=1}^{n+1}(x_i-{\hat{\mu }}_n)(x_i-{\hat{\mu }}_n{)}^t& =\sum _{i=1}^n(x_i-{\hat{\mu }}_n)(x_i-{\hat{\mu }}_n{)}^t+(x_{n+1}-{\hat{\mu }}_n)(x_{n+1}-{\hat{\mu }}_n{)}^t\\ & =(n-1)C_n+(x_{n+1}-{\hat{\mu }}_n)(x_{n+1}-{\hat{\mu }}_n{)}^t\\ & =(n-1)C_n+[(n+1){\hat{\mu }}_{n+1}-n{\hat{\mu }}_n-{\hat{\mu }}_n][(n+1){\hat{\mu }}_{n+1}-n{\hat{\mu }}_n-{\hat{\mu }}_n{]}^t\\ & =(n-1)C_n+(n+1{)}^2({\hat{\mu }}_{n+1}-{\hat{\mu }}_n)({\hat{\mu }}_{n+1}-{\hat{\mu }}_n{)}^t\end{array}$$

$$\sum _{i=1}^{n+1}({\hat{\mu }}_i-{\hat{\mu }}_{n+1})({\hat{\mu }}_i-{\hat{\mu }}_{n+1}{)}^t=(n+1)({\hat{\mu }}_n-{\hat{\mu }}_{n+1})({\hat{\mu }}_n-{\hat{\mu }}_{n+1}{)}^t$$

$$\begin{array}{rl}\sum _{i=1}^{n+1}\left(x_i-{\hat{\mu }}_n\right){\left({\hat{\mu }}_n-{\hat{\mu }}_{n-1}\right)}^t& =[\sum _{i=1}^{n+1}\left(x_i+{\hat{u}}_n\right)]{\left({\hat{\mu }}_n-{\hat{\mu }}_{n-1}\right)}^t\\ & =(x_1+...+x_n+x_{n+1}-(n+1){\hat{\mu }}_n){\left({\hat{\mu }}_n-{\hat{\mu }}_{n+1}\right)}^t\\ & =(n{\hat{\mu }}_n+x_{n+1}-(n+1){\hat{\mu }}_n){\left({\hat{\mu }}_n-{\hat{\mu }}_{n+1}\right)}^t\\ & =(-{\hat{\mu }}_n+x_{n+1}){\left({\hat{\mu }}_n-{\hat{\mu }}_{n+1}\right)}^t\\ & =(-{\hat{\mu }}_n+(n+1){\hat{\mu }}_{n+1}-n{\hat{\mu }}_n){\left({\hat{\mu }}_n-{\hat{\mu }}_{n+1}\right)}^t\\ & =(n+1)({\hat{\mu }}_{n+1}-{\hat{\mu }}_n){\left({\hat{u}}_n-{\hat{\mu }}_{n+1}\right)}^t\\ & =-(n+1)\left({\hat{\mu }}_n-{\hat{\mu }}_{n+1}\right){\left({\hat{\mu }}_n-{\hat{\mu }}_{n+1}\right)}^t\end{array}$$
类似的，有
$$\sum _{i=1}^{n+1}({\hat{\mu }}_n-{\hat{\mu }}_{n+1})(x_i-{\hat{\mu }}_n{)}^t=-(n+1)\left({\hat{\mu }}_n-{\hat{\mu }}_{n+1}\right){\left({\hat{\mu }}_n-{\hat{\mu }}_{n+1}\right)}^t$$
带回原式有
$$n{C}_{n+1}=(n-1)C_n+n(n+1)({\hat{\mu }}_{n+1}-{\hat{\mu }}_n)({\hat{\mu }}_{n+1}-{\hat{\mu }}_n{)}^t$$
又由
$$(n+1)({\hat{\mu }}_{n+1}-{\hat{\mu }}_n)=\frac{1}{n+1}(x_{n+1}-{\hat{\mu }}_n)$$
化简得
$$C_{n+1}=\frac{n-1}{n}{C}_n+\frac{1}{n+1}\left(x_{n+1}-{\hat{\mu }}_n\right){\left(x_{n+1}-{\hat{\mu }}_n\right)}^t$$