first, i need to know which software do you use

or you write it by yourself

by the way, is the guy you mean granger?

there should be two variable before your correction term

one is cointegration vector, the other one is error adjust term, look at the original model

if the cointegration vector is insignificant, that means there is no cointegration

if the error adjust term is insignificant, that means it is perfect cointegrated

I do the paper by EViews.

"格兰杰表示定理" is "Granger Representation Theorem".

fj102 said that "if the error adjust term is insignificant, that means it is perfect cointegrated", i understand that "perfect cointegrated" means there is no ECM for the variables which are perfect cointegrated. Right or wrong ?

I donot think you need to say there is no ECM, because basically ECM and cointegration are equivalent.

What you need to describe is that the adjustment of past error to long run equilibrium is near zero.Hence there is no error correct term.

"near zero" doesn't mean "no"!

I got it!

Thanks a lot!

我觉得这种情况比较可能。而且

“我的情况是这样的:两个变量的图形显示为几乎平行向右上角延伸,这种"几乎平行"是不是属于易丹辉老师所说的"超稳定状态"呢?”

也说明两个变量本身就是TSP，即都是I（0），所以，不显著很正常。

If the variable is I(0), how can it has a trend?

如果一个序更是趋势平稳（Trend Stable Process),我们仍认为它是I(0)的吧。但一阶差分平稳（Differenc Stable Process)，则我们认为它是I(1)的，所以，TSP仍然是I(0)，即这种趋势不是“漂移”的，这也是在ADF检验中可以加入时间趋势项的原因，做ADF之前最好先看一下图形，看看究竟有没有Trend。

could you let me know which book describe it

if there is trend stable process, we regard it as I(0)

You have to understand the basic concept of stationary and non stationary

usually stationary means mean or variance-covariance is not time dependent

if you have trend, then it is no longer belongs to I(0)

there is also strict definition about stationary using density function, we wont care about that.

平稳的概念我略知一点啊。

一般说的平稳指的是“宽平稳”，即均值与自协方差不随时间变化的随机过程

而严格平稳则是联合概率分布不变的随机过程

宽平稳与严平稳没有必然联系，严平稳不是宽平稳的充分条件，也不是必要条件

then can you tell me the difference between the trend stable process and thd first order difference stable process? and , the ADF test D(y)=c+b*t+p*y_t-1+sigma(D(y_t-1....))，is it only and only if when there no b*t term, the stable process can be called I(0), otherwise, the no such concept I(0) ? I'm a little confused

[此贴子已经被作者于2006-2-2 21:10:40编辑过]

我没有看过一本书上说过趋势平稳是I(0)的，不过，关于平稳、趋势平稳与差分平稳，则看过Gujarati的Basic Econometrics上面有过论述。其实他的意思是将TSP等同于I(0)，

I(0)有三种，

即如果用ADF形式为 D(y)=c+p*y_t-1+sigma(D(y_t-i....))，不含趋势项

一种为 D(y)=b*t+p*y_t-1+sigma(D(y_t-1....))，含有趋势项

另一种为D(y)=c+b*t+p*y_t-1+sigma(D(y_t-i....))，c 和 b*t　项都含有

三种如果p＝0成立，则都不是平稳的，否则，都可视为平稳

不过，好象按标准的定义，平稳指的是均值和自协方差平稳，含有趋势项的显然不符合均值不变的条件。

[此贴子已经被作者于2006-2-2 21:08:51编辑过]

I find it in Hamilton book as well

I think you have to read it carefully

it is a DF regression model

one of the key null hypothesis, true process, is the coefficient of time t is zero.

After estimation, the a*t term, a is expected to be zero

this is a test regression model, experts just add the time trend to enhence their results

anyway, read the book again

it is just a basic concept

As to this ECM model,I think there does exist some problem, especially in explaining real economic activities, like Hendry said in our seminar, granger's ECM misleading our current research.

james davidson seems make some progress on it, however, still follow granger's routine. Cannot make real improvement.