概率统计——样本方差

  令样本的变量分布$X$的均值为$\mu$，方差为${\sigma }^2$。对其随机样本 { x 1 , . . . , x n } \{x_1, ..., x_n\} {x1​,...,xn​}，其样本方差为$$s^2=1/(n-1)\sum _{i=1}^n(x_i-\bar{x}{)}^2$$其中$n-1$令$s^2$是样本的无偏估计，即$$E(s^2)={\sigma }^2$$究其缘由，我们定义假设样本方差为$${\hat{s}}^2=1/n\sum _{i=1}^n(x_i-\bar{x}{)}^2$$则有$$\begin{array}{rl}{\hat{s}}^2& =1/n\sum _{i=1}^n(x_i-\bar{x}{)}^2\\ & =1/n\sum _{i=1}^n((x_i-\mu )-(\bar{x}-\mu ){)}^2\\ & =1/n\sum _{i=1}^n(x_i-\mu {)}^2+1/n\sum _{i=1}^n(\bar{x}-\mu {)}^2-2/n\sum _{i=1}^n(x_i-\mu )(\bar{x}-\mu )\\ & =1/n\sum _{i=1}^n(x_i-\mu {)}^2+(\bar{x}-\mu {)}^2-2(\bar{x}-\mu )(\bar{x}-\mu )\\ & =1/n\sum _{i=1}^n(x_i-\mu {)}^2-(\bar{x}-\mu {)}^2\\ & \le 1/n\sum _{i=1}^n(x_i-\mu {)}^2\end{array}$$而$1/n{\sum }_{i=1}^n(x_i-\mu {)}^2$是真实方差，即$$\hat{s}\le s$$当样本均值与分布均值一致时取等号，即该式非无偏估计。
  为了得到无偏估计，考虑
$$\begin{array}{rl}E[{\hat{s}}^2]& =E[1/n\sum _{i=1}^n(x_i-\bar{x}{)}^2]\\ & =E[1/n\sum _{i=1}^n{x}_i^2+1/n\sum _{i=1}^n{\bar{x}}^2-2/n\sum _{i=1}^n{x}_i\bar{x}]\\ & =E[1/n\sum _{i=1}^n{x}_i^2+{\bar{x}}^2-2\bar{x}/n\sum _{i=1}^n{x}_i]\\ & =E[1/n\sum _{i=1}^n{x}_i^2+{\bar{x}}^2-2{\bar{x}}^2]\\ & =E[1/n\sum _{i=1}^n{x}_i^2-{\bar{x}}^2]\\ & =E[1/n\sum _{i=1}^n{x}_i^2]-E[{\bar{x}}^2]\\ & =E[x_i^2]-E[{\bar{x}}^2]\end{array}$$考虑方差公式$$DX=E[X^2]-E{X}^2$$有均值的期望$$E[\bar{x}]=\bar{x}$$以及均值的方差$$D[\bar{x}]=D[1/n\sum _{i=1}^n{x}_i]=1/n^2\sum _{i=1}^{n}D[x_i]=1/nD[x_i]$$故
$$\begin{array}{rl}E[{\hat{s}}^2]& =E[x_i^2]-E[{\bar{x}}^2]\\ & =D[x_i]+E[x_i{]}^2-D[\bar{x}]-E[\bar{x}{]}^2\\ & =D[x_i]-1/nD[x_i]\\ & =(n-1)/nD[x_i]\end{array}$$因此为了令$$E[s^2]=D[x_i]$$有$$s^2=n/(n-1){\hat{s}}^2$$即$$s^2=n/(n-1)[1/n\sum _{i=1}^n(x_i-\bar{x}{)}^2]=1/(n-1)\sum _{i=1}^n(x_i-\bar{x}{)}^2$$证毕。