[求助]协整滞后选择的问题？

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

我在作协正过程中发现，滞后值越大，AIC和SC准则的值就越小；相应J协整得到的关系就越多；但可有数据就减少。

请问大家，J协整检验对于滞后项该如何取？是取滞后最大值，还是尽可能的小一点呢？

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2 回帖

关键词：协整检验 最大值 AIC 尽可能 滞后项 选择

其实，做协整研究 都是不科学的， 每个不同的规则， 都会导致不同的滞后， 当然结果就不一样了， 我不知道这些做法还有什么意义，反正最后都能牵强自圆其说的解释， 当然了，这些方法本身没有什么毛病，只不过 临界值 都是在渐进大样本的 条件下使用的， 然而我们40来个数据， 照用不误。

然而我们40来个数据， 照用不误: interesting, don't they know what is asymptotics? ^_^

谁规定 "大样本" 就一定要样本大呢？！ Says who?

Hint：  你是从 undifferenced data 的 VAR 开始决定 lag terms 吗？

谁规定 "大样本" 就一定要样本大呢？！ Says who?

Come on! if that is so, then why people replace chi squared test statistics by F one to have a better small sample property.

Why in nearly every paper in "Econometric Theory"-like that pursuing to give asymptotic test statictis will give simulation result for relevant sample size to have an idea about the size and power of the test.

Says who?: there are many!

gemini69: for example

Tse 2000 J of Econometrics, his LM test has an empirical power of 50% for sample size 300, and 90% for sample size 1000. it over-rejects the H0 for 300 and is well-sized for 1000.

Of course the sample size should be reasonably large since the test statistic is based on CLT. well you can say that 30 is infinity. But without enough number of obs, no usful information is available.

[此贴子已经被作者于2006-2-13 3:12:57编辑过]

gemini69:

maybe you can recommand me a paper in the core English Econometric journal that deals with cointegration using only 40 obs.

以下是引用zhaosweden在2006-2-13 3:14:00的发言：

gemini69:

maybe you can recommand me a paper in the core English Econometric journal that deals with cointegration using only 40 obs.

Well，

1、"渐近" ：随着样本渐大的性质，并不等同样本就要很大才会有的性质，它不是恒等式；何况，多少的样本数才是大？多少的样本数才是小，或是以多少样本数区分有限与渐近？ 你也拜托一下，这是质的观念，不是量的标准！

2、这跟收敛速度有关吧！

3、 40个年资料，两个变量，每一个就有20年的资料，严不严谨是一回事，叁考价值多不多是另一回事。

4、不需要给 paper吧，这根本是基本观念。不然你翻一下任何一本基础的概率或是计量教科书，有没有那个作者，给你拍胸保证，样本个数40的估计式，不管什麽状况，钢定、铁定不符合渐近性质。

随便摘录几段话：

幼稚园级： (Basic Econometrics 4E, Gujarati, pp903)

" It often happens that an estimator does not satisfy one or more of the desirable statistical properties

in small samples. But as the sample size increases indefinitely, the estimator possesses several desirable

statistical properties. These properties are known as the large-sample, or asymptotic, properties."

幼稚园级： (Introductory Econometrics 1st Edition, Wooldridge, pp68)

" Asymptotic analysis involves approximating the features of the sampling distribution of an estimator.

These approximations depend on the size of the sample. Unfortunately, we are necessarily limited in

what we can say about how “large” a sample size is needed for asymptotic analysis to be appropriate;

this depends on the underlying population distribution. But large sample approximations have been

known to work well for sample sizes as small as n ＝ 20. "

初级： (Econometric Theory and Methods_Davidson & MacKinnon, pp145)

"Asymptotic theory is concerned with the distributions of estimators and test statistics as the sample size n

tends to infinity. It often allows us to obtain simple results which provide useful approximations even when the

sample size is far from infinite."

计量学者： 台湾中研院院士， PhD, UCSD (Prof. White 的学生)

" 许多书都强调样本规模必须大过 30 (或 50) 才够，但事实上这个问题是没有答案的。

因为：有时样本规模小于 30 时，参数与估计值便非常接近。有时即使样本规模大到 3000，

估计值仍不会接近参数。 所以我们只能说 n 越大，估计值愈有可能非常接近真实参数。

(请参阅课本214页，图7.1)"

Usually depend on your sample, if your sample distribution is consistent with population distribution,

that is no problem, just as above, work well for small sample

if the sample distribution is not consistent with population distribution,

unfortunately, you have to use large sample to approximate it

the only problem is how we know whehter the sample distribution is consistent or not

usually we choose large sample to get convergence in case there is consistence problem