计量经济学-最小二乘估计量

计量经济学-最小二乘估计量无偏性及有效性证明&区间估计与假设检验

已知：

线性
$Y_t=\alpha +\beta X_t+{\mu }_t$
同方差
$Var({\mu }_i)=Var({\mu }_j)={\sigma }^2$
无关性
$E({\mu }_i{\mu }_j)=0$
外生性
$E({\mu }_i\mid X)=0$
正态分布
${\mu }_i\sim N(0,{\sigma }^2)$

$\hat{\beta }=\frac{{\sum }_{i=1}^N(X_i-\overline{X})(Y_i-\overline{Y})}{{\sum }_{i=1}^N(X_i-\overline{X}{)}^2}$

无偏性证明：

$\begin{array}{rl}E({\hat{\beta }}_{ols}\mid X)& =E(\frac{{\sum }_{i=1}^N(X_i-\overline{X})(Y_i-\overline{Y})}{{\sum }_{i=1}^N(X_i-\overline{X}{)}^2}\mid X)\\ & \\ & =\frac{{\sum }_{i=1}^N[E(X_i-\overline{X})(Y_i-\overline{Y})\mid X]}{{\sum }_{i=1}^N(X_i-\overline{X}{)}^2}\\ & \\ & =\frac{{\sum }_{i=1}^N(X_i-\overline{X})E(Y_i-\overline{Y}\mid X)}{{\sum }_{i=1}^N(X_i-\overline{X}{)}^2}\\ & \\ & =\frac{{\sum }_{i=1}^N(X_i-\overline{X})[E(Y_i\mid X)-E(\overline{Y}\mid X)]}{{\sum }_{i=1}^N(X_i-\overline{X}{)}^2}\\ & \\ & =\frac{{\sum }_{i=1}^N(X_i-\overline{X})[E(\alpha +\beta X_i+{\mu }_i\mid X)-E(\frac{1}{N}{\sum }_{i=1}^N{Y}_i\mid X)]}{{\sum }_{i=1}^N(X_i-\overline{X}{)}^2}\\ & \\ & =\frac{{\sum }_{i=1}^N(X_i-\overline{X})(\alpha +\beta X_i-\alpha -\beta \overline{X})}{{\sum }_{i=1}^N(X_i-\overline{X}{)}^2}\\ & \\ & =\frac{\beta {\sum }_{i=1}^N(X_i-\overline{X}{)}^2}{{\sum }_{i=1}^N(X_i-\overline{X}{)}^2}\\ & \\ & =\beta \end{array}$

$\because E({\hat{\beta }}_{ols}\mid X)=\beta$

$\therefore E(E({\hat{\beta }}_{ols}\mid X))=E(\beta )$

$\therefore E({\hat{\beta }}_{ols})=\beta$

有效性证明：

即证明最小二乘估计量的方差小于其他线性无偏估计量的方差

最小二乘估计方差：

$\begin{array}{rl}{\hat{\beta }}_{ols}& =\frac{{\sum }_{i=1}^N(X_i-\overline{X})(Y_i-\overline{Y})}{{\sum }_{i=1}^N(X_i-\overline{X}{)}^2}\\ & \\ & =\frac{{\sum }_{i=1}^N(X_i-\overline{X})Y_i}{{\sum }_{i=1}^N(X_i-\overline{X}{)}^2}-\frac{{\sum }_{i=1}^N(X_i-\overline{X})\overline{Y}}{{\sum }_{i=1}^N(X_i-\overline{X}{)}^2}\end{array}$

$\begin{array}{rl}\because \sum _{i=1}^N(X_i-\overline{X})\overline{Y}& =\sum _{i=1}^N(X_i-\overline{X})\overline{Y}\\ & \\ & =0\end{array}$
$\begin{array}{rl}\therefore {\hat{\beta }}_{ols}& =\frac{{\sum }_{i=1}^N(X_i-\overline{X})Y_i}{{\sum }_{i=1}^N(X_i-\overline{X}{)}^2}\\ & \\ & =\frac{{\sum }_{i=1}^N(X_i-\overline{X})(\alpha +\beta X_i+{\mu }_i)}{{\sum }_{i=1}^N(X_i-\overline{X}{)}^2}\\ & \\ & =\frac{\beta {\sum }_{i=1}^N{X}_i(X_i-\overline{X})+{\sum }_{i=1}^N{\mu }_i(X_i-\overline{X})}{{\sum }_{i=1}^N(X_i-\overline{X}{)}^2}\\ & \\ & =\frac{\beta {\sum }_{i=1}^N(X_i-\overline{X}{)}^2+\beta {\sum }_{i=1}^N(X_i-\overline{X})\overline{X}+{\sum }_{i=1}^N{\mu }_i(X_i-\overline{X})}{{\sum }_{i=1}^N(X_i-\overline{X}{)}^2}\\ & \\ & =\beta +\frac{{\sum }_{i=1}^N{\mu }_i(X_i-\overline{X})}{{\sum }_{i=1}^N(X_i-\overline{X}{)}^2}\end{array}$

$\therefore {\hat{\beta }}_{ols}-\beta =\frac{{\sum }_{i=1}^N{\mu }_i(X_i-\overline{X})}{{\sum }_{i=1}^N(X_i-\overline{X}{)}^2}$

V a r ( β ^ o l s ) = E { [ β ^ o l s − E ( β ^ o l s ) ] 2 } = E [ ( β ^ o l s − β ) 2 ] = E { [ ∑ i = 1 N μ i ( X i − X ‾ ) ∑ i = 1 N ( X i − X ‾ ) 2 ] 2 } = E { [ ∑ i = 1 N ( X i − X ‾ ) μ i ∑ i = 1 N ( X i − X ‾ ) 2 ] 2 } = E { ∑ i = 1 N ( X i − X ‾ ) 2 μ i 2 + ∑ i = 1 , i ≠ j N ( X i − X ‾ ) ( X j − X ‾ ) μ i μ j [ ∑ i = 1 N ( X i − X ‾ ) 2 ] 2 } = ∑ i = 1 N ( X i − X ‾ ) 2 E ( μ i 2 ) + ∑ i = 1 , i ≠ j N ( X i − X ‾ ) ( X j − X ‾ ) E ( μ i μ j ) [ ∑ i = 1 N ( X i − X ‾ ) 2 ] 2 = σ 2 ∑ i = 1 N ( X i − X ‾ ) 2 [ ∑ i = 1 N ( X i − X ‾ ) 2 ] 2 = σ 2 ∑ i = 1 N ( X i − X ‾ ) 2 \begin{aligned} Var(\hat{\beta}_{ols}) & = E\{[\hat{\beta}_{ols} - E(\hat{\beta}_{ols})]^2\}\\ \\ & = E[(\hat{\beta}_{ols} - \beta)^2]\\ \\ & = E\{[\cfrac{\sum_{i = 1}^N\mu_i(X_i - \overline X)}{\sum_{i = 1}^N(X_i - \overline X)^2}]^2\}\\ \\ & = E\{[\cfrac{\sum_{i = 1}^N(X_i-\overline X)\mu_i}{\sum_{i = 1}^N(X_i - \overline X)^2}]^2\}\\ \\ & = E\{\cfrac{\sum_{i = 1}^N(X_i - \overline X)^2\mu_i^2 + \sum_{i = 1,i\ne j}^N(X_i - \overline X)(X_j - \overline X)\mu_i\mu_j}{[\sum_{i = 1}^N(X_i - \overline X)^2]^2}\}\\ \\ & = \cfrac{\sum_{i = 1}^N(X_i - \overline X)^2E(\mu_i^2) + \sum_{i = 1,i\ne j}^N(X_i - \overline X)(X_j - \overline X)E(\mu_i\mu_j)}{[\sum_{i = 1}^N(X_i - \overline X)^2]^2}\\ \\ & = \cfrac{\sigma^2\sum_{i = 1}^N(X_i - \overline X)^2}{[\sum_{i = 1}^N(X_i - \overline X)^2]^2}\\ \\ & = \cfrac{\sigma^2}{\sum_{i = 1}^N(X_i - \overline X)^2}\\ \end{aligned} Var(β^​ols​)​=E{[β^​ols​−E(β^​ols​)]2}=E[(β^​ols​−β)2]=E{[∑i=1N​(Xi​−X)2∑i=1N​μi​(Xi​−X)​]2}=E{[∑i=1N​(Xi​−X)2∑i=1N​(Xi​−X)μi​​]2}=E{[∑i=1N​(Xi​−X)2]2∑i=1N​(Xi​−X)2μi2​+∑i=1,i=jN​(Xi​−X)(Xj​−X)μi​μj​​}=[∑i=1N​(Xi​−X)2]2∑i=1N​(Xi​−X)2E(μi2​)+∑i=1,i=jN​(Xi​−X)(Xj​−X)E(μi​μj​)​=[∑i=1N​(Xi​−X)2]2σ2∑i=1N​(Xi​−X)2​=∑i=1N​(Xi​−X)2σ2​​

其他线性无偏估计方差：

假设：$\beta$的任一线性估计量为${\sum }_{i=1}^N{\omega }_i{Y}_i$。
因此：
$\begin{array}{rl}E({\hat{\beta }}^{\ast })& =\sum _{i=1}^N{\omega }_{i}E(Y_i)\\ & \\ & =\sum _{i=1}^N{\omega }_{i}E(\alpha +\beta X_i)\\ & \\ & =\alpha \sum _{i=1}^N{\omega }_i+\beta \sum _{i=1}^N{\omega }_i{X}_i\end{array}$
$\because \beta$为无偏估计量
$\therefore E({\hat{\beta }}^{\ast })=\beta$

$\therefore$ $\alpha {\sum }_{i=1}^N{\omega }_i=0$且${\sum }_{i=1}^N{\omega }_i{X}_i=1$

$\begin{array}{rl}Var({\hat{\beta }}^{\ast })& =Var(\sum _{i=1}^N{\omega }_i{Y}_i)\\ & \\ & =\sum _{i=1}^N{\omega }_i^{2}Var(Y_i)\\ & \\ & =\sum _{i=1}^N{\omega }_i^{2}Var(\alpha +\beta X_i+{\mu }_i)\\ & \\ & =\sum _{i=1}^N{\omega }_i^{2}Var({\mu }_i)\\ & \\ & ={\sigma }^2\sum _{i=1}^N{\omega }_i^2\end{array}$

$\because$ 柯西不等式

$\begin{array}{rl}\therefore \sum _{i=1}^N{\omega }_i^2\sum _{i=1}^N(X_i-\overline{X}{)}^2& \ge [\sum _{i=1}^N[{\omega }_i\times (X_i-\overline{X})]{]}^2\\ & \\ & \ge (\sum _{i=1}^N{\omega }_i{X}_i-\overline{X}\sum _{i=1}^N{\omega }_i{)}^2\\ & \\ & \ge 1\end{array}$

$\therefore$ ${\sum }_{i=1}^N{\omega }_i^2\ge \frac{1}{{\sum }_{i=1}^N(X_i-\overline{X}{)}^2}$

$\therefore$ ${\sigma }^2{\sum }_{i=1}^N{\omega }_i^2\ge \frac{{\sigma }^2}{{\sum }_{i=1}^N(X_i-\overline{X}{)}^2}$

即$Var({\hat{\beta }}^{\ast })\ge Var({\hat{\beta }}_{ols})$

置信区间

$\because Var({\hat{\beta }}_{ols})=\frac{{\sigma }^2}{{\sum }_{i=1}^N(X_i-\overline{X}{)}^2}$
$\therefore Se({\hat{\beta }}_{ols})=\frac{\sigma }{\sqrt{{\sum }_{i=1}^N(X_i-\overline{X}{)}^2}}$
$\because \sigma$为未知量

$\therefore$需要用残差$e$来估计$\sigma$

${\hat{\sigma }}^2=\frac{{\sum }_{i=1}^N{e}_i^2}{n-2}$

$\because$ $\{\begin{array}{ll}{Y}_i& =\hat{\alpha }+\hat{\beta }{X}_i+e_i\\ & \\ \overline{Y}& =\hat{\alpha }+\hat{\beta }\overline{X}\end{array}$

$\therefore$ $Y_i-\overline{Y}=\hat{\beta }(X_i-\overline{X})+e_i$

$\therefore$ $e_i=(Y_i-\overline{Y})-\hat{\beta }(X_i-\overline{X})$

$\therefore$ $\begin{array}{rl}\sum _{i=1}^N{e}_i^2& =\sum _{i=1}^N[(Y_i-\overline{Y})-\hat{\beta }(X_i-\overline{X}){]}^2\\ & \\ & =\sum _{i=1}^N(Y_i-\overline{Y}{)}^2-2\hat{\beta }\sum _{i=1}^N(X_i-\overline{X})(Y_i-\overline{Y})+{\hat{\beta }}^2\sum _{i=1}^N(X_i-\overline{X}{)}^2\\ & \\ & =\sum _{i=1}^N(Y_i-\overline{Y}{)}^2+\hat{\beta }[\hat{\beta }\sum _{i=1}^N(X_i-\overline{X}{)}^2-2\sum _{i=1}^N(X_i-\overline{X})(Y_i-\overline{Y})]\end{array}$

$\because$ $\hat{\beta }=\frac{{\sum }_{i=1}^N(X_i-\overline{X})(Y_i-\overline{Y})}{{\sum }_{i=1}^N(X_i-\overline{X}{)}^2}$

$\therefore$ $\begin{array}{rl}\sum _{i=1}^N{e}_i& =\sum _{i=1}^N(Y_i-\overline{Y}{)}^2+\hat{\beta }[\frac{{\sum }_{i=1}^N(X_i-\overline{X})(Y_i-\overline{Y})}{{\sum }_{i=1}^N(X_i-\overline{X}{)}^2}\cdot \sum _{i=1}^N(X_i-\overline{X}{)}^2-2\sum _{i=1}^N(X_i-\overline{X})(Y_i-\overline{Y})]\\ & =\sum _{i=1}^N(Y_i-\overline{Y}{)}^2-\hat{\beta }\sum _{i=1}^N(X_i-\overline{X})(Y_i-\overline{Y})\end{array}$

$\therefore$ ${\hat{\sigma }}^2=\frac{{\sum }_{i=1}^N(Y_i-\overline{Y}{)}^2-\hat{\beta }{\sum }_{i=1}^N(X_i-\overline{X})(Y_i-\overline{Y})}{n-2}$

$\therefore$ $Var({\hat{\beta }}_{ols})=\frac{{\sum }_{i=1}^N(Y_i-\overline{Y}{)}^2-\hat{\beta }{\sum }_{i=1}^N(X_i-\overline{X})(Y_i-\overline{Y})}{(n-2){\sum }_{i=1}^N(X_i-\overline{X}{)}^2}$

$\therefore$ $Se({\hat{\beta }}_{ols})=\sqrt{\frac{{\sum }_{i=1}^N(Y_i-\overline{Y}{)}^2-\hat{\beta }{\sum }_{i=1}^N(X_i-\overline{X})(Y_i-\overline{Y})}{(n-2){\sum }_{i=1}^N(X_i-\overline{X}{)}^2}}$

假设检验