Estimating Granger causality for many variables.

[Amir Moayeri]

Hi. I have a list of 5 dependent variables (y1…y5) and a list of 5 independent variables (x1…x5).
I want to test the causality from each ‘x’ to its corresponding ‘y’ by this command
var y x, lags(1/3)
vargranger
and store the variable names and the related p-values in three new columns in this form
y1 x1 p-value1
y2 x2 p-value2
and so on.
Is it possible to do that in Stata?

[George Ford]

Sure. Could use collect or store results in a matrix.

matrix won't store names, but could use the number on y and x and then add those when you display the results of the matrix.

[Amir Moayeri]

I wrote this code to access the p-value. How can I make a list of p-values by adding the values in each step?

var y1 x1, lags(1/3)
vargranger
 
mata
stats = st_matrix("r(gstats)")
st_numscalar("pvalue", stats[2,3])
end
di pvalue

[George Ford]

matrix R = J(5,2,.)
forv i = 1/5 {
qui var y`i' x`i', lags(1/3)
qui vargranger
matrix R[`i',1] = `i'
matrix R[`i',2] = el(r(gstats),1,3)
di "Model: y`i' <- x`i'   Prob = " %5.3f R[`i',2]
}
matrix list R