结果就是得出错误答案。

2# soccy

但是  如果解释变量和误差项是不相关的话  按道理说得出的coefficient就是无偏的呀，，，额 不好意思  我比较弱，，，

解释变量和误差项不相关只是需要满足的诸多条件之一。

"但是  如果解释变量和误差项是不相关的话  按道理说得出的coefficient就是无偏的呀" True, but if the error covariance still has struction pattern, then it is NOT an efficient estimator.

"额 不好意思  我比较弱" -- You already know a lot. Work hard!!!

In term of assumtions in OLS, the E(error * x) = 0 is the most important one. Under the model is correctly specified, the heteroscedastic and serial correlation only affect efficiency of estimation. But when high serial correlation exists, one has to deal with it.It will improve your estimation.

Test serial correlation is a simple procedure assume that you are talking about autocorrelation in error term , I am sure EViews should have it.