This post suggests that we use eigen values to determine the half life of reverting to mean :
I have read in many places that the alpha matrix represents the speed of adjustment.
I read in another place that the eigenvalue is approximately equal to the "alpha" parameter.
Here is a counter example :-
The eigen value is .0038. None of the values in the alpha (loading matrix) are close to this.
Give a VECM with a alpha matrix, how do I compute the half life ? In case we have only ONE regression we do ln(2)/alpha.
My query is how do we go from the alpha matrix to the half life in the case when alpha is a matrix ?