Automatic Lag Selection

[camilogong]

Hi,
I am trying to run a routine to perform the best lag selection for a regression. When I run it it doesn't pick the best lag, but rather takes the last lag available.
The routine looks like this

for !i=0 to 12
!r2=0
%fcistring= " "
%fcibest=" "
if !i=0 then %fcistring= "fci"+@str({!j}{!k}{!l}{!m}{!n}{!o}{!p})
else %fcistring= "fci"+@str({!j}{!k}{!l}{!m}{!n}{!o}{!p})+ "("+@str(-{!i})+")"
endif
smpl 2005m09 @last-12
equation r{!j}{!k}{!l}{!m}{!n}{!o}{!p}
r{!j}{!k}{!l}{!m}{!n}{!o}{!p}.ls g c {%arstring} {%fcistring}
if r{!j}{!k}{!l}{!m}{!n}{!o}{!p}.@rbar2>!r2 then
!r2=r{!j}{!k}{!l}{!m}{!n}{!o}{!p}.@rbar2
%fcibest =%fcistring
r{!j}{!k}{!l}{!m}{!n}{!o}{!p}.ls g c {%arstring} {%fcibest}
else delete r{!j}{!k}{!l}{!m}{!n}{!o}{!p}
endif
next


Can someone give me a hand? I really appreciate it

[EViews Gareth]

I don't see anything that jumps out at me as being wrong.

[camilogong]

I know, that is what is driving me completely crazy.
Could you run the routine and see if you get any different results?
I attach the program and the xls
Rgds

[EViews Gareth]

You're resetting !r2 inside the loop :D

Move the !r2=0 to before the for loop.

Also, from an econometric point of view, you might want to check you're using the same number of observations in each equation.

[camilogong]

Actually, setting the !r2 outside of the loop makes does not change matters all that much, as it still does not do the lag selection
What puzzles me is that I am currently runninng an identical program for another country and that one seems to be runnning (and doing the lag selection) fine (attached).
You are right that I need to set a fixed sample for the selection criteria to make sense.
Any clue would be greatly appreciated.