How do I choose the best rolling regression?

[jm13]

Hi

I have run 5 rolling regressions. This was an example of my programme if needed:

workfile thesisnsesp200
!window = 20
!step = 1
!length = @obsrange
system sys09
sys09.append spr=c(1)
sys09.append nser=c(2)
!nrolls = @round((!length- !window)/!step)
matrix(9, !nrolls) coefmat
matrix(9, !nrolls) tstats
!j=0
for !i = 1 to !length- !window+1- !step step !step
!j=!j+1
smpl @first+!i-1 @first+!i+!window-2
sys09.arch(b,h) @diagbekk c(fullrank) arch(1) garch(1)
colplace(coefmat, sys09.@coefs, !j)
colplace(tstats, sys09.@tstats, !j)
next

show coefmat
show tstats

Is there even a way to choose the best one? Do I have to add in a line/s asking for each AIC, SIC etc and then would I average it out - if so how would I programme that in? Or do I just take the rolling regression with the most observations as the 20-, 30- and 50-week rolling regressions have much fewer observations than the 100 and 200 week rolling regressions?

Please help. I would be grateful for anyone's advice and assistance.

Thanks

Jason