Thank you for your attention. I am still learning this ECM and VECM.
Some people point out to me several issues in VECM:
1. Small sample bias.
2. Data should be integrated in order(1).
<script>
open data10-5
loop foreach j nc c ct ctt gls --quiet
loop foreach i calwage uswage --quiet
adf 4 $i --$j --test-down --quiet
lt_$i$j=$test
lp_$i$j=$pvalue
adf 4 $i --$j --difference --test-down --quiet
dt_$i$j=$test
dp_$i$j=$pvalue
end loop
end loop
loop foreach j nc c ct ctt gls --quiet
printf "\nADF $j\n \t\t%10s \t%12s \n", "Level",
"Difference"
printf "\t\t%5s [%6s] %5s [%6s]\n", "tvalue", "pvalue",
"tvalue", "pvalue"
loop foreach i calwage uswage --quiet
printf "$i \t%6.3f [%6.4f] %6.3f [%6.4f]\n", lt_$i$j, lp_$i$j, dt_$i$j,
dp_$i$j
end loop
end loop
</script>
I cannot find how to access the lag order in ADF test. Is there any accessor for the
actual lag order used?
Considering small sample data and ADF test result, can we use VECM?
Yuniarto Hadiwibowo
----- Original Message ----
From: Allin Cottrell <cottrell(a)wfu.edu>
To: Gretl list <gretl-users(a)lists.wfu.edu>
Sent: Tuesday, December 9, 2008 2:21:40 AM
Subject: Re: [Gretl-users] ECM
<script>
open data10-5
coint2 2 calwage uswage
vecm 2 1 calwage uswage
restrict
a[2] = 0
end restrict --verbose
</script>
The "restrict" block tests the hypothesis that the US wage does
not error-correct in response to perturbations to the California
wage, which is implicitly assumed in the single-equation
specification; this hypothsis is rejected at the 5 percent level.
Allin Cottrell