Re: [Gretl-users] ADF result
On Thu, 4 Dec 2008, Sven Schreiber wrote:
It's absolutely clear in the help and manual, no doubt. That lag
orders always refer to the dependent variable in unit-root
contexts however is _not_ a universal convention.
I surely don't mind trying to make things as clear as possible,
but I'm not certain I'm understanding the issue here (and may be
missing something).
Isn't it the case that in an ADF test the "lag order" always
refers to lags of the dependent variable *in the test regression*?
This "dependent variable" will typically be the first difference
of the original variable of interest, y, though it might be the
second difference if one had already concluded that y follows a
random walk and one wanted to check for stationarity in \delta y.
But would one ever want to regress \delta y on y(t-1) plus further
lags of y, i.e. y(t-2) etc., rather than adding lags of \delta y?
Well, of course one might want to do that, but would it be an ADF
test?
Allin.