Rolling Impulse Response Functions and FEVD Calculation

[deneda]

Hi,
For my study, I need to calculate the forecast error variance decompositions of my variables which I estimate through rolling VAR. Inspired by one of the former posts, I think I managed to get the impulse response functions out of rolling VAR. However, I need the forecast error variance decompositions (FEVD) of these variables whereas what I write merely gives me the graph of the irfs. How can I create what I want?
The codes I have used till now are as follows. I am assuming they are right, but of course they might be wrong.
My VAR estimates stock returns of U.S, Frankfurt and London stock markets simultaneously.
I would be really glad if you could help.

scalar window = 200

for !i = 1 to window
var var.ls 1 2 dus dger dlse
freeze(roliin) var.impulse(12, se=a) dus @dger @dlse
var.makeresids error_dus error_dger error_dlse

[EViews Glenn]

http://www.eviews.com/help/helpintro.ht ... 23ww175710