VAR model

[Elie]

Hi ,

i need a help

i conducted a VAR model using e-view program
after choosing the suitable lag length i re estimate the VAR ,so i got
estimated coefficient
Standard error
t-statistic

the q is : how can i determinate which (estimated coefficient) is a significant ?

[vinux]

You can look at t-statistic. Roughly if it is less than -2 or greater than 2 then it is significant.
i.e. significant if |t-statisitc| >2.

[Elie]

Thanks Vinux
but actually i don't believe this is a true,
1- i think you mean 2%
2- i want to test if it is significant at 1% and 5%..
so do you have any good idea about how to conduct that ?
i appreciate your help
regards ...

[Elie]

also i believe that the df of the sample play a role in the test ? don't you ?

[vinux]

Thanks Vinux
but actually i don't believe this is a true,
1- i think you mean 2%
2- i want to test if it is significant at 1% and 5%..
so do you have any good idea about how to conduct that ?
i appreciate your help
regards ...

Strictly speaking under null hypothesis t-statistics follows student's t distribution with error df ( this is depending on the number of parameters involved in the model.
But for a large sample t-statistics is almost same as z statistics. So you can compare t-statistic with z critical values. For example for 5%, z critical value is 1.96 ( approximately 2).

[Elie]

that's OK , so what is the critical value for 1%
and i'm sorry but i have a big gap in the econometric ,but 1.96 in % or just 1.96 ??

[Elie]

It should be 2.326 , but that 's actually confusing me,

In an old study, it considered the 0.0009 for example significant at 1% ? why ?
if there is any guide in this forum help me with the VAR issue ,i will appreciate that

[vinux]

It is not %

P[|Z| >1.96] = 0.05 # So 1.96 corresponds to 5% (you need to take the both tails)
which is same as P[Z>1.96] = 0.025
and
P[|Z| >2.56] = 0.01 # So 2.56 corresponds to 1%

Usually the alternative hypothesis is not equal to, this lead to two tail test.

[Elie]

which is same as P[Z>1.96] = 0.025

mmm i didn't get this point

lol i know now you are saying

but why in the old study it ,the researcher considered 0.0009 as significant at 1%

[Elie]

Dears, if there is any threads about Vector Autogressive model in the forum i appreciate if you tell me about it?

[vinux]

mmm i didn't get this point

lol i know now you are saying

but why in the old study it ,the researcher considered 0.0009 as significant at 1%

Regarding the calculation, i hope you know Z ( standard normal) is a symmetric distribution.
P[|Z|<1.96] =0.95. => P[-1.96 <Z<1.96] = 2*P[0<Z<1.96] = 0.95
P[0<Z<1.96] = 0.95/2 & P[Z>0]=0.5 => P[Z>1.96] = 0.025.

This would be easy if you can visualize the area of the tails.

but why in the old study it ,the researcher considered 0.0009 as significant at 1%

I don't understand this part. What is 0.0009 value?

[Dason]

My guess is that the p-value was .0009 - not the t-statistic.

[Elie]

first of all i really appreciate your help and patient

Regarding the calculation, i hope you know Z ( standard normal) is a symmetric distribution.
P[|Z|<1.96] =0.95. => P[-1.96 <Z<1.96] = 2*P[0<Z<1.96] = 0.95
P[0<Z<1.96] = 0.95/2 & P[Z>0]=0.5 => P[Z>1.96] = 0.025.

This would be easy if you can visualize the area of the tails.

OK so i have to compare t-stat with the value 1.96 if the |t-stat|>1.96 => the estimated coefficient is significant in 5% ,is it right ?
when i tried this in the example in Gujarti the results make a sense ,but Gujarti make it significant at 10% (p851)
and he accepted a lower value such as 1.87699 ,is there any explanation ?

[Elie]

My guess is that the p-value was .0009 - not the t-statistic.

actually it is the t -statistic

the results is as below :
Standard errors in () & t-statistics in [].
3.09E−05
(0.03198)
[0.00097]***

*VAR system has been estimated with 4 lags according to Akaike Information Selection Criterion with oil as an essential reference.
*** indicates significance at 1% level

so ????