El 22/02/12 13:44, alexkakashi(a)libero.it escribió:
Dear community of gretl,
I prefer writing a new post about my problem. I'm interested in testing
Granger causality from X to Y in a trivariate VAR. The critical values are
found using bootstrap based on residuals. In the first stage I consider only
the equation
ols Y const Y(-1) Y(-2) X(-1) X(-2) Z(-1) Z(-2)
Under the null hypothesis I replace the series Y using the residuals of this
equation and the coefficient estimated under the null. The test statistics is
large and observing the p-value (small) I reject the null hypothesis of no-
causality.
I have also considered the VAR model and, in the bootstrap method, I have
replace every equation of the VAR. I have used this commands:
r=158 # number of observations
system method=sur
equation Y const Y(-1) Y(-2) X(-1) X(-2) Z(-1) Z(-2)
equation X const Y(-1) Y(-2) X(-1) X(-2) Z(-1) Z(-2)
equation Z const Y(-1) Y(-2) X(-1) X(-2) Z(-1) Z(-2)
end system
...
Alessandro,
In a system of equations with the same explanatory variables for all the
equations, the SUR method gives exactly the same results as OLS, and OLS
of the system is exactly equivalent to OLS in each individual equation.
This means that you may use the 'var' command in gretl and, unless you
suspect a strong non-normality in your data, you may use the Granger
causality tests provided by the 'var' command (they are the F tests at
the end of each equation).
--
Ignacio Diaz-Emparanza
DEPARTAMENTO DE ECONOMÍA APLICADA III (ECONOMETRÍA Y ESTADÍSTICA)
UPV/EHU Avda. Lehendakari Aguirre, 83 | 48015 BILBAO
T.: +34 946013732 | F.: +34 946013754
www.ea3.ehu.es