On Tue, 13 Jul 2010, Summers, Peter wrote:
> Ah, my example wasn't very well though-out. You're right that
> with series like {gdp cons inv}, a common trend/cointegrating
> vector would be more interesting. But suppose I have a list of
> commodity price series like {oil gold copper}. A panel unit root
> test might be more relevant there (at least according to my
> referees!).
If you really want to fake a panel unit root test:
scalar T = $nobs
matrix pan = {oil} | {gold} | {copper}
scalar n = rows(pan)
nulldata n --preserve
series fake = pan
setobs T 1:01 --stacked-time-series
adf 4 fake # or whatever
Allin Cottrell
