[求助]广义差分法处理模型一阶自相关后为什么T检验变得不通过呢(具体模型见正文),该怎么处理?各位大侠指教!

是 否 +2 论坛币

k人 参与回答

经管之家送您一份

应届毕业生专属福利!

求职就业群

赵安豆老师微信：zhaoandou666

经管之家联合CDA

送您一个全额奖学金名额~ !

立即领取

感谢您参与论坛问题回答

经管之家送您两个论坛币！

+2 论坛币

各位高人,小弟在用EVIEWS做一元回归的时候发现模型存在一阶自相关,于是用广义差分法处理,但回归结果让我很意外,原先都通过T的常项和自变量系数变得都不通过,但拟合优度提高不少,做原点回归的广义差分法则可以通过T检验,但拟合优度降低,具体见下面,请问这种情况该怎么处理?望各们前辈不吝赐教,先行谢过!

原始回归:Y = 1.781817308 + 1.020360123*X

T检验P=0.0054 T检验P= 0.0000 R-squared=0.916566 模型存在显著一阶序列相关.

广义差分:Y= 86.06351508 - 0.09304939932*X+ [AR(1)=0.9987221206]

T检验P=0.9016 T检验P=0.1048 T检验P=0.0000 R-squared=0.997596 模型显著不存在自相关

原点回归差分:Y = 1.23094033*X + [AR(1)=0.7852002674]

T检验P=0.0000 T检验P=0.0024 R-squared=0.911165 模型自相关检验够玄,均不显著

扫码加我 拉你入群

请注明：姓名-公司-职位

以便审核进群资格，未注明则拒绝

分享0 收藏1 回帖

关键词：广义差分法 怎么处理 各位大侠 广义差分 自相关 模型 大侠 自相关 差分法 正文

You'd better to read carefuly chapter 12 in Wooldridge "introductory Econometrics" 2nd Edn

Since beta1 changed sign, the huge difference between OLS and Cochrane-Orcutt or Prais-Winston is always an indication of problem. choice: try differenced series; work with OLS and Newy and West standard Error., And others...

Thanks a lot!can you tell me some details?

i think you should make a unit root test on Y, X and the residuals

test cointegration

I did do the cointegration test firstly, and the regression is not furious, please tell me more details,thaks!

Thanks a lot!can you tell me some details

----------

When it comes to the details, we need equations and formula! So You can try to read Wooldridge, he explains it very well, especially regarding the intuitions.

来看看那

肿么都整英语

哈哈，先看单位根和协整