I have a SVAR model that is estimated with the vars package. All the variables are logged and first differenced. I understand that the impulse response functions (IRF) shows how much a 1 standard deviation shock in the impulse variable (let's call it x) is associated with the response variable (let's call it y).
Since the variables are not in levels, what would be the correct way to interpret an impulse response coefficient of 0.28?