计量经济学08.ppt

Appendices Assume our model is: The least squares estimator for β2 is: where 8A Properties of the Least Squares Estimator Eq. 8A.1 Since E(ei) = 0, we get: so β2 is unbiased 8A Properties of the Least Squares Estimator But the least squares standard errors are incorrect under heteroskedasticity: 8A Properties of the Least Squares Estimator Eq. 8A.2 If the variances are all the same, then Eq. 8A.2 simplifies so that: 8A Properties of the Least Squares Estimator Eq. 8A.3 Recall that: and assume that our objective is to test H0: α2 = α3 = … = αS = 0 against the alternative that at least one αs, for s = 2, ... , S, is nonzero 8B LaGrange Multiplier Tests for Heteroskedasticity Eq. 8B.1 The F-statistic is: where 8B LaGrange Multiplier Tests for Heteroskedasticity Eq. 8B.2 We can modify Eq. 8B.2 Note that: Next, note that: Substituting, we get: 8B LaGrange Multiplier Tests for Heteroskedasticity Eq. 8B.3 Eq. 8B.4 Eq. 8B.5 If it is assumed that ei is normally distributed, then: and: Further: 8B LaGrange Multiplier Tests for Heteroskedasticity Eq. 8B.6 Using Eq. 8B.6, we reject a null hypothesis of homoskedasticity when the χ2-value is greater than a critical value from the χ2(S – 1) distribution. 8B LaGrange Multiplier Tests for Heteroskedasticity Also, if we set: We can get: 8B LaGrange Multiplier Tests for Heteroskedasticity Eq. 8B.7 Eq. 8B.8 At a 5% significance level, a null hypothesis of homoskedasticity is rejected when exceeds the critical value χ2(0.95, S – 1) 8B LaGrange Multiplier Tests for Heteroskedasticity 另一种形式的异方差表现为：样本可以划分为两个或两个以上的群组，每个群组具有不同的误差方差。 8.4 Generalized Least Squares: Known Form of Variance 8.4.12 Grouped Data 大部分软件具有加权最小二乘或广义最小二乘的选项。 8.4 Generalized Least Squares: Known Form of Variance 8.4.1b Weighted Least Squares 对工资方差，有: with 8.4 Generalized Least Squares: Known Form of Variance 8.4.1b Weighted Least Squares Eq. 8.33a Eq. 8.33b 基于 不同的误差方差，分别获得最小二乘估计： 但是，我们得到β2 和