异方差与序列相关性练习.docVIP

一、异方差检验与修正 （一）建立初始回归模型 相关命令： data x y scat x y ls y c x 模型一： Dependent Variable: Y Method: Least Squares Date: 10/23/14 Time: 10:46 Sample: 1 20 Included observations: 20 Variable Coefficient Std. Error t-Statistic Prob.?? C 272.3635 159.6773 1.705713 0.1053 X 0.755125 0.023316 32.38690 0.0000 R-squared 0.983129 ????Mean dependent var 5199.515 Adjusted R-squared 0.982192 ????S.D. dependent var 1625.275 S.E. of regression 216.8900 ????Akaike info criterion 13.69130 Sum squared resid 846743.0 ????Schwarz criterion 13.79087 Log likelihood -134.9130 ????F-statistic 1048.912 Durbin-Watson stat 1.301684 ????Prob(F-statistic) 0.000000 （二）异方差的四种检验方法及其分析 右击resid选择Object Copy，输入e得到初始回归模型的残差序列； 1. 图示法：scat x e^2 2. 模型检验法：ls e^2 c x Dependent Variable: E^2 Method: Least Squares Date: 10/23/14 Time: 10:52 Sample: 1 20 Included observations: 20 Variable Coefficient Std. Error t-Statistic Prob.?? C -65281.66 21544.58 -3.030073 0.0072 X 16.49344 3.145895 5.242843 0.0001 R-squared 0.604286 ????Mean dependent var 42337.15 Adjusted R-squared 0.582302 ????S.D. dependent var 45279.67 S.E. of regression 29264.05 ????Akaike info criterion 23.50075 Sum squared resid 1.54E+10 ????Schwarz criterion 23.60032 Log likelihood -233.0075 ????F-statistic 27.48740 Durbin-Watson stat 1.029463 ????Prob(F-statistic) 0.000055 3. GQ假设检验法 首先，点击工具按钮proc选择sort current page，输入X，按升序排序；去掉中间约n/4个样本点，然后对前后两个子样本分别进行回归； 子样本模型一： Dependent Variable: Y Method: Least Squares Date: 10/23/14 Time: 10:57 Sample: 1 8 Included observations: 8 Variable Coefficient Std. Error t-Statistic Prob.?? C 1277.161 1540.604 0.829000 0.4388 X 0.554126 0.311432 1.779287 0.1255 R-squared 0.345397 ????Mean dependent var 4016.814 Adjusted R-squared 0.236296 ????S.D. dependent var 166.1712 S.E. of regression 145.2172 ????Akaike