'1、做一个方程
equation ols.ls gdp c cs inv gov_net
show ols
'2、做white异方差检验(（如果是no cross则如下，否则为ols.white(c））
ols.white

White’s test is a test of the null hypothesis of no heteroskedasticity against heteroskedasticity
of some unknown general form. The test statistic is computed by an auxiliary regression,
where we regress the squared residuals on all possible (nonredundant) cross
products of the regressors.

Now this is the result, What can we know about it?

White Heteroskedasticity Test:

F-statistic 13.08147 Prob. F(6,184) 0.000000
Obs*R-squared 57.11238 Prob. Chi-Square(6) 0.000000


Test Equation:
Dependent Variable: RESID^2
Method: Least Squares
Date: 05/14/07 Time: 09:55
Sample: 1947Q2 1994Q4
Included observations: 191

Variable Coefficient Std. Error t-Statistic Prob.

C 19314.49 2233.775 8.646569 0.0000
CS -22.08585 3.275450 -6.742845 0.0000
CS^2 0.005651 0.000938 6.024419 0.0000
INV 25.57363 8.010134 3.192659 0.0017
INV^2 -0.045871 0.009317 -4.923321 0.0000
GOV_NET -10.84981 13.49588 -0.803935 0.4225
GOV_NET^2 -0.048491 0.044609 -1.087023 0.2784

R-squared 0.299018 Mean dependent var 2342.796
Adjusted R-squared 0.276160 S.D. dependent var 3146.576
S.E. of regression 2677.069 Akaike info criterion 18.65879
Sum squared resid 1.32E+09 Schwarz criterion 18.77799
Log likelihood -1774.915 F-statistic 13.08147
Durbin-Watson stat 0.524501 Prob(F-statistic) 0.000000

First,From F-statistic and Obs*R-squared, we konw taht the test results refuse the null hypothesis, namely there is heteroskedasticity.

Second,the following equation is used to test heteroskedasticity, which tell you how the test run!

可以用white检验来做，由于样本不大，所以最好选无交叉项。因为如果选择有交叉项，在多元回归中，辅助回归方程中有太多解释变量，从而会使得自由度大大减少。你可以参考李子奈的《计量经济学》（第二版）第99页的说明，也可看高铁梅《计量经济分析方法与建模：Eviews应用及实例》P117.高铁梅在做white检验时，样本只有30个，也不是很大哦。

还有就是G—Q检验，但这种方法要排序，在多元回归中无法判断哪个解释变量引起异方差。

white检验不是要大样本吗？

回答：F-statistic 可对应精确分布；Obs*R-squared服从卡方分布，可对应近似分布。这就是应对大小样本的一个非常粗浅的理解。

再说，何为大小样本？统计定义模糊！实质上这是相对于4楼所指的自由度而言！我在二楼提供的是no cross的white检验，目的就是为了减少自由度损失。

30在统计中不是认为是大样本和小样本的一个最低分界嘛