Granger Causality - VAR/VECM/?
Posted: Mon Jun 28, 2010 5:45 am
Hello!
I am sorry if my question might seem stupid, but I was wondering how to best construct my granger causality test. I am investigating two time series, over a 5 year period with daily info and I also know that these time series are cointegrated. How do I know proceed with my causality test?
1. I thought about eventually simply taking the time series themselves, opening them as a group and conducting granger causlity tests in this way...Does this show me the granger causlity of the price development (my time series are price development series)
2. If i have cointegration I could also conduct the Granger Causality test on the basis of a VECM. In this case, for the lag interval do I simply take the lag interval used for the johansen test? or do I take this lag interval minus 1 (as VECM demands differenced data) and if, what happens if my original lag length was only 1 1? or do the lags need to be established in any other way?
3.if i conduct the granger causality with VECM what exactly am i comparing? the prie development? the models?
thank you in advance i really appreciate any help!!!
best regards,
I am sorry if my question might seem stupid, but I was wondering how to best construct my granger causality test. I am investigating two time series, over a 5 year period with daily info and I also know that these time series are cointegrated. How do I know proceed with my causality test?
1. I thought about eventually simply taking the time series themselves, opening them as a group and conducting granger causlity tests in this way...Does this show me the granger causlity of the price development (my time series are price development series)
2. If i have cointegration I could also conduct the Granger Causality test on the basis of a VECM. In this case, for the lag interval do I simply take the lag interval used for the johansen test? or do I take this lag interval minus 1 (as VECM demands differenced data) and if, what happens if my original lag length was only 1 1? or do the lags need to be established in any other way?
3.if i conduct the granger causality with VECM what exactly am i comparing? the prie development? the models?
thank you in advance i really appreciate any help!!!
best regards,