Granger causality lag length

[moojf]

HI

If I performe Granger causality test on Eviews between two variables.

How can I decide the lag length neded

Thanks

[random_access]

i cannot remember how it looks like....is there any AIC or SC values there?
there is another way to select the right lag but first let me know about the selection criteria

[moojf]

In Granger causality there is no criteria

It just ask you for the lag length.

What i found that i have to choose the larger lag that match the theory
is that right?

[random_access]

it depends on the data frequency.
select the variables and run VAR
when the estimation output comes out go to view and then lag structure and then lag length criteria and then use 30 if u r using daily data and 12 if monthly and check the criteria.....it will give u few criteria....choose one and use it for causality
hope it help

[moojf]

Thank you sir

[moojf]

it depends on the data frequency.
select the variables and run VAR
when the estimation output comes out go to view and then lag structure and then lag length criteria and then use 30 if u r using daily data and 12 if monthly and check the criteria.....it will give u few criteria....choose one and use it for causality
hope it help

What about if I have a yearly observations 40 years?
Thanks

[random_access]

i think it should be allright

[moojf]

i think it should be allright

sorry what you mean?

I didn`t get that?

And thanks again for your help

[random_access]

if u r using yearly data and u have 40 years. then u can use one year.....usually they do not go beyond.
so u can conduct Granger causality using one year back
hope it helps

[moojf]

That`s fair enough

Thnaks

[moojf]

Hi Random

If I have Quartrely data 136 observation, what should i use?
Thanks

[DOC12]

I am trying to carryout pair wise granger causality for about 16 series, I have tried using the VAR method to select the optimal lag length. It seems that the different criteria except SC (which always chooses zero) just automatically select the max lag lenth chosen, I tried up to 40 lags and I got the same thing. I know the consensus is that the optimal lag length for daily data is thirty and monthly data is twelve, but what about weekly data. Thanks

[gr031]

if u r using yearly data and u have 40 years. then u can use one year.....usually they do not go beyond.
so u can conduct Granger causality using one year back
hope it helps

I have yearly data with 13 observations. What lag length should I choose? Thanks!

[may.k]

I also would like to know the lag length for 30 annual observation. when I use lag 4, it was found uni direction. is lag length 4 appropriate for 30 annual observation?