1、控制变量的异方差问题应该考虑，这些在stata里面完全可以通过在命令后加上“，r”来得到稳健标准差来解决；
2、共线性变量不一定必须删除。如果多重共线性不形象你研究的问题的结果则无需关注，如果造成不良影响再处理；
3、修正之后再检验一下就是了，如果你的修正方法正确，结果肯定不会存在异方差问题。

1. 应该考虑control variable的异方差性吗

For continuous variable,

Plot residual against control variable to see if the residual fluctuation has patterns , say control variable is bigger and fluctuation is bigger.

For a categorical variable,

plot a histogram of residual against control variable.

2.是否应该在检验之前先删除存在共线性的变量呢? Yes. Usually the condistion number above 20 needs to take a look.

3. 修正之后还要再进行异方差性检验吗? Yes.

sorry i do not know about condition number related multicollinearity. i used the correlation matrix. a correlation coefficient between two explanatory variables greater than 0.8 or 0.9 in absolute value indicates a strong linear association and a potentially harmful collinear relationship.

i have one pair with the value as high as around 0.99...but which variable should i delete or should i delete both of them?

about heteroskedasticity, i used the white.s approximate estimator for the variance of the least squares estimator 修正异方差,但是,eviews做出来以后,数据并没有发现变化, 我用来检测异方差的方法是 the Goldfeld-quandt这个方法是要把数据分成平均的两份,然后等等....可是white修正完了以后,我不知道该怎么再测试. 书上根本没提到这个问题. 或许我应该用别的检测方法? 或是别的修正方法? 或者,我应该用gls?

啥玩意啊,看不懂