我想问下，Granger因果关系中X和Y的滞后阶数p和q，这两者是不是可以不相等的？

可以呀

呵呵,我的意思不是抨击文章作者的水平低,他们都是高人,我根本没资格对他们品头论足

但是我觉得根据IC来定阶数实在太武断.我曾经比较过一些股指和汇率的数据,发现AIC,SBIC,MSPE给出的结论都不相同,而且有时会差得相当远.在决定阶数时很头疼.(不知大家有什么建议没有.)如果是做Model的话其实问题不大,顶多是准确度的问题.但Granger因果测试对阶数很敏感,几乎到了大是大非的程度,如果就这样粗糙地定下阶数那最后的结果也就显得些苍白.

此外,我对金融计量经济学在实际中的应用有些担忧,特别是线性模型.我算出来的SR(预测回报正负符号的正确率)一般都是50%左右,根本就是瞎猜.非线性模型的表现会好一些,但准确率还是不能让人很惊喜.相对而言,物理学的流体力学,飞行器制导等方向比较难学,但至少科学家现在已经能把人送上太空.而金融计量经济学这个领域却很少能见到这样振奋人心的成果.

可能又说多了几句,也许说错了不少,请大家批

hi, Netdagger,

I don't what happened, I couldn't send you email, I hence post my reply here and I suppose you will see it soon.

thank you for your question, I have never think of it before.

accordingto the definition of r-square, r-square is for knowing howwell themodel explain the data set, but in deciding the lag-length, we wannaknow the power of new-added lag value in explaining the present value,hence, it is meaningless to consider r-square in decidinglag length.actually, r-square should be growing as the increase of laglengthsuppose there is no colinearity and autocorelation caused bynew-addedlag value, you can porve this if you want to get down to thisquestion,alternatively, you also can find solution in many books. ifyou want tosay r-square paly a role in deciding the lag length, Isuggest you checkthe change of the first difference of r-square, maybeit contains someinformation about how good the model is, but in myview, it doesn't makesense.

above is my view, I cann't assure you that my interpretation isright, if you have your interpretation, please let me know.

Hi, oldman

Thanks for your msg, I will try it in the future and report the result to u.

Many

Hi, oldman

Thanks for your msg, I will try it in the future and report the result to u.

Many

但是这些准则在eviews上怎么看得到呢？

受益匪浅~~~~~~

最近也在做这方面的文章，头大。。。

两个准则来判断，值较小时的阶数.....................

那个granger分析怎么看两个准则，具体操作是什么啊

请问Granger的滞后期和VAR的是一样的吗？

请问Granger的滞后期和VAR的是一样的吗？

一般来说，Granger检验是基于VAR建模的基础上的，所以滞后期在一般情况下是一样的。但是。。。，如果你使用T-Y Granger Test, 是有可能不同的。

这都是我十年前的回复啦。。。 祝好