Hello,
I test for granger causality and i got the following results:
y1 granger cause y2
y2 does not granger cause y1
In this situation, can I estimate a VAR model in order to forecast y1 and y2 at time t?
Thank you in advance!
Ideally, VAR modelling shall confirm and follow Granger's Causality.
Yes I know but I need to forecast one variable in function of the lags of the other one. So in this case I cannot do this?
Hi Vanessa. You cannot use your current VAR for an interdependent forecast. You may use y1 to forecast y2, but not the other way around. If you attempt to make a forecast under these circumstances, most programs will make the calculations without problem, but you would only be making forecasts of y1 based on its own lagged values according to the lag-length that you select. Be sure that this conforms to your research needs...
Dear Ms. Hanna,
Granger Causality is a causality a relational test while VAR is a Regression Method which comes under the chapter on Correlations and Regressions. Besides, Granger Causality does not require a parametric model necessarily while VAR does. So the two are not the same . Therefore you can report the results of both tests with more informational value to the user. That's what I think.
Soumitra K. Mallick
for RHMHM School
You said that VAR requires a parametric model. What do you mean? Which law it follows?
Vanessa:
In your situation, y1 is exogenous with respect to y2. You can estimate a modified VAR where y1 depends only on its own lagged values, while y2 depends on lagged y2 and lagged y1 (I'm going to assume that y1 and y2 are stationary and not co-integrated).
- Badges
- Science topic