1、问题描述:假设有一批独立同分布的样本{
x
i
,
i
=
1
,
2
,
3...
,
n
−
1
,
n
{x_i,i=1,2,3...,n-1,n}
xi,i=1,2,3...,n−1,n}。
标准方差的公式为:
S
2
=
1
n
−
1
∑
i
=
1
n
(
x
i
−
x
‾
)
2
S^{2} = \frac{1}{n-1}\sum_{i=1}^{n}(x_i - \overline{x})^{2}
S2=n−11i=1∑n(xi−x)2
,其中
x
‾
=
1
n
∑
i
=
1
n
x
i
\overline{x}=\frac{1}{n}\sum_{i=1}^{n}x_i
x=n1∑i=1nxi;
另外假设均值 E ( x i ) = μ E(x_i) = \mu E(xi)=μ,方差 D ( x i ) = σ 2 D(x_i) = \sigma^{2} D(xi)=σ2;
证明:标准方差是方差的无偏估计。
【分析】:要想证明标准方差是方差的无偏估计,只需证明 E ( S 2 ) = σ 2 = D ( x i ) E(S^2) = \sigma^{2} = D(x_i) E(S2)=σ2=D(xi),其中比较麻烦的地方是 x i ∗ x ‾ x_i * \overline{x} xi∗x,所以要把 x ‾ \overline{x} x中的 x i x_i xi分离处理。
2、证明开始:
E
(
S
2
)
=
E
(
1
n
−
1
∑
i
=
1
n
(
x
i
−
x
‾
)
2
)
=
1
n
−
1
E
(
∑
i
=
1
n
(
x
i
−
1
n
x
i
−
1
n
∑
j
≠
i
,
j
=
1
n
x
j
)
2
)
=
1
n
−
1
E
(
∑
i
=
1
n
(
(
n
−
1
)
2
n
2
x
i
2
−
2
(
n
−
1
)
n
2
x
i
∗
∑
j
≠
i
,
j
=
1
n
x
j
+
1
n
2
(
∑
j
≠
i
,
j
=
1
n
x
j
)
2
)
)
\begin{aligned} E(S^{2}) &= E(\frac{1}{n-1}\sum_{i=1}^{n}(x_i - \overline{x})^{2}) \\ &= \frac{1}{n-1}E(\sum_{i=1}^{n}(x_i - \frac{1}{n}x_i - \frac{1}{n}\sum_{j\neq{i},j=1}^{n}x_j)^{2})\\ &= \frac{1}{n-1}E(\sum_{i=1}^{n}(\frac{(n-1)^2}{n^2}x_i^{2} - \frac{2(n-1)}{n^{2}}x_i *\sum_{j\neq{i},j=1}^{n}x_j + \frac{1}{n^2}(\sum_{j\neq{i},j=1}^{n}x_j)^{2})) \\ \end{aligned}
E(S2)=E(n−11i=1∑n(xi−x)2)=n−11E(i=1∑n(xi−n1xi−n1j=i,j=1∑nxj)2)=n−11E(i=1∑n(n2(n−1)2xi2−n22(n−1)xi∗j=i,j=1∑nxj+n21(j=i,j=1∑nxj)2))
下面将
E
(
⋅
)
E(\centerdot)
E(⋅)中的三部分分别计算,按照从左往右的顺序,系数
1
n
−
1
\frac{1}{n-1}
n−11暂时不带,最后算完结果一起带入。
第一部分:
E
(
∑
i
=
1
n
(
(
n
−
1
)
2
n
2
)
x
i
2
)
=
(
n
−
1
)
2
n
2
∑
i
=
1
n
E
(
x
i
2
)
=
(
n
−
1
)
2
n
2
∑
i
=
1
n
(
E
2
(
x
i
)
+
D
(
x
i
)
)
=
(
n
−
1
)
2
n
2
∑
i
=
1
n
(
μ
2
+
σ
2
)
=
(
n
−
1
)
2
n
(
μ
2
+
σ
2
)
\begin{aligned} E(\sum_{i=1}^{n}(\frac{(n-1)^2}{n^2})x_i^{2}) &= \frac{(n-1)^2}{n^2}\sum_{i=1}^{n}E(x_i^{2}) \\ &= \frac{(n-1)^2}{n^2}\sum_{i=1}^{n}(E^2(x_i) + D(x_i)) \\ &= \frac{(n-1)^2}{n^2}\sum_{i=1}^{n}(\mu^{2} + \sigma^{2}) \\ &= \frac{(n-1)^2}{n}(\mu^{2} + \sigma^{2}) \end{aligned}
E(i=1∑n(n2(n−1)2)xi2)=n2(n−1)2i=1∑nE(xi2)=n2(n−1)2i=1∑n(E2(xi)+D(xi))=n2(n−1)2i=1∑n(μ2+σ2)=n(n−1)2(μ2+σ2)
第二部分:
E
(
∑
i
=
1
n
(
2
(
n
−
1
)
n
2
x
i
∗
∑
j
≠
i
,
j
=
1
n
x
j
)
)
=
2
(
n
−
1
)
n
2
∑
i
=
1
n
E
(
x
i
∗
∑
j
≠
i
,
j
=
1
n
x
j
)
)
=
2
(
n
−
1
)
n
2
∑
i
=
1
n
(
E
(
x
i
)
∗
∑
j
≠
i
,
j
=
1
n
E
(
x
j
)
)
=
2
(
n
−
1
)
n
2
∑
i
=
1
n
(
μ
∗
(
n
−
1
)
μ
)
=
2
(
n
−
1
)
n
2
∗
n
∗
(
n
−
1
)
∗
μ
2
=
2
(
n
−
1
)
2
∗
μ
2
n
\begin{aligned} E(\sum_{i=1}^{n}( \frac{2(n-1)}{n^{2}}x_i *\sum_{j\neq{i},j=1}^{n}x_j)) &= \frac{2(n-1)}{n^{2}}\sum_{i=1}^{n}E(x_i *\sum_{j\neq{i},j=1}^{n}x_j)) \\ &= \frac{2(n-1)}{n^{2}}\sum_{i=1}^{n}(E(x_i) * \sum_{j\neq{i},j=1}^{n}E(x_j)) \\ &= \frac{2(n-1)}{n^{2}}\sum_{i=1}^{n}(\mu * (n-1)\mu) \\ &= \frac{2(n-1)}{n^{2}} * n*(n-1)*\mu^{2} \\ & = \frac{2(n-1)^{2}*\mu^{2}}{n} \end{aligned}
E(i=1∑n(n22(n−1)xi∗j=i,j=1∑nxj))=n22(n−1)i=1∑nE(xi∗j=i,j=1∑nxj))=n22(n−1)i=1∑n(E(xi)∗j=i,j=1∑nE(xj))=n22(n−1)i=1∑n(μ∗(n−1)μ)=n22(n−1)∗n∗(n−1)∗μ2=n2(n−1)2∗μ2
第三部分:
E
(
∑
i
=
1
n
(
1
n
2
(
∑
j
≠
i
,
j
=
1
n
x
j
)
2
)
)
=
1
n
2
∑
i
=
1
n
E
(
(
(
∑
j
≠
i
,
j
=
1
n
x
j
)
2
)
)
=
1
n
2
∑
i
=
1
n
(
E
2
(
∑
j
≠
i
,
j
=
1
n
x
j
)
+
D
(
∑
j
≠
i
,
j
=
1
n
x
j
)
)
=
1
n
2
∑
i
=
1
n
(
(
∑
j
≠
i
,
j
=
1
n
E
2
(
x
j
)
)
+
∑
j
≠
i
,
j
=
1
n
D
(
x
j
)
)
=
1
n
2
∑
i
=
1
n
(
(
(
∑
j
≠
i
,
j
=
1
n
μ
)
2
)
+
∑
j
≠
i
,
j
=
1
n
σ
2
)
=
1
n
2
∑
i
=
1
n
(
(
n
−
1
)
2
∗
μ
2
+
(
n
−
1
)
∗
σ
2
)
=
(
n
−
1
)
2
n
2
∗
n
∗
μ
2
+
n
−
1
n
2
∗
n
σ
2
=
(
n
−
1
)
2
n
∗
μ
2
+
n
−
1
n
σ
2
\begin{aligned} E(\sum_{i=1}^{n}(\frac{1}{n^2}(\sum_{j\neq{i},j=1}^{n}x_j)^{2})) &= \frac{1}{n^2}\sum_{i=1}^{n}E(((\sum_{j\neq{i},j=1}^{n}x_j)^{2})) \\ &= \frac{1}{n^2}\sum_{i=1}^{n}(E^{2}(\sum_{j\neq{i},j=1}^{n}x_j) +D(\sum_{j\neq{i},j=1}^{n}x_j)) \\ &= \frac{1}{n^2}\sum_{i=1}^{n}((\sum_{j\neq{i},j=1}^{n}E^{2}(x_j))+\sum_{j\neq{i},j=1}^{n}D(x_j)) \\ &= \frac{1}{n^2}\sum_{i=1}^{n}(((\sum_{j\neq{i},j=1}^{n}\mu)^{2})+\sum_{j\neq{i},j=1}^{n}\sigma^{2}) \\ &= \frac{1}{n^2}\sum_{i=1}^{n}((n-1)^{2}*\mu^2+(n-1)*\sigma^2) \\ &= \frac{(n-1)^2}{n^2} *n *\mu^2+\frac{n-1}{n^2} *n\sigma^2\\ &= \frac{(n-1)^2}{n} *\mu^2+\frac{n-1}{n} \sigma^2 \end{aligned}
E(i=1∑n(n21(j=i,j=1∑nxj)2))=n21i=1∑nE(((j=i,j=1∑nxj)2))=n21i=1∑n(E2(j=i,j=1∑nxj)+D(j=i,j=1∑nxj))=n21i=1∑n((j=i,j=1∑nE2(xj))+j=i,j=1∑nD(xj))=n21i=1∑n(((j=i,j=1∑nμ)2)+j=i,j=1∑nσ2)=n21i=1∑n((n−1)2∗μ2+(n−1)∗σ2)=n2(n−1)2∗n∗μ2+n2n−1∗nσ2=n(n−1)2∗μ2+nn−1σ2
终于到了相加的时候了,将1-3部分的内容带上系数相加如下:
(
n
−
1
)
E
(
⋅
)
=
(
n
−
1
)
2
n
(
μ
2
+
σ
2
)
−
2
(
n
−
1
)
2
∗
μ
2
n
+
(
n
−
1
)
2
n
∗
μ
2
+
n
−
1
n
σ
2
=
(
n
−
1
)
2
n
σ
2
+
n
−
1
n
σ
2
=
n
2
−
n
n
σ
2
=
(
n
−
1
)
σ
2
\begin{aligned} (n-1)E(\centerdot) & =\frac{(n-1)^2}{n}(\mu^{2} + \sigma^{2}) \ - \frac{2(n-1)^{2}*\mu^{2}}{n} + \frac{(n-1)^2}{n} *\mu^2+\frac{n-1}{n} \sigma^2 \\ &= \frac{(n-1)^2}{n}\sigma^{2} + \frac{n-1}{n} \sigma^2 \\ &= \frac{n^2-n}{n}\sigma^{2} \\ &= (n-1)\sigma^{2} \end{aligned}
(n−1)E(⋅)=n(n−1)2(μ2+σ2) −n2(n−1)2∗μ2+n(n−1)2∗μ2+nn−1σ2=n(n−1)2σ2+nn−1σ2=nn2−nσ2=(n−1)σ2
所以
E
(
⋅
)
=
σ
2
E(\centerdot) = \sigma^2
E(⋅)=σ2
得证。
注:标准方差前面的系数
1
n
−
1
\frac{1}{n-1}
n−11是为了让其是方差的无偏估计而加上的。