Re: [Gretl-users] autoregrssive estimation
On Mon, 12 Feb 2007, Javier García wrote:
I've done the exercise adding 1 2 3 4 5 6 7 8 9 and 1 2 3 4 5 6
7 9 in the lags' list, so in independt variable I've included no
variables. My data set is weekly, but I've tried with yearly
data and the result is the same (error with data). I've got
other dubt: which is the difference between add 3 independt
variables' lags and add 3 lags in AR lags' list?
Let me take the second doubt first. The "AR lags list" refers to
lags of the error term, as opposed to lags of the dependent
variable. The model is
(1) y(t) = X(t)*B + u(t)
(2) u(t) = r_1*u(t-1) * r_2*u(t-2) + ... + e(t)
By selecting the "AR lags" you are choosing which r_i values to
allow.
That's why you get a "Data error" (the message should be more
informative!) when you use this command with no independent
variables specified (not even a constant, or any lags of the
dependent variable). The trouble is you have a null regression
function. That is,
u(t) = y(t) - X(t)*B
but X(t)*B is undefined since X is empty.
Allin Cottrell