Em 18 de março de 2011 Allin escreveu:

On Thu, 17 Mar 2011, Henrique Andrade wrote:

> I would suggest a change in the ADF test (through the GUI). When Gretl asks
> for the lag order I think it would be nice to put a default value for the
> lag order (also leaving the option "test down from maximum lag order")
> selected by default.
>
> Looking at EViews (ADF test)  and R (Phillips-Perron test), I'd discovered
> that they use the following default, as suggested by Schwert (Tests for Unit
> Roots: A Monte Carlo Investigation, Journal of Business & Economic
> Statistics, 1989, p.9):
>
> *scalar adf_lag = int(12*($nobs/100)^(1/4))*

I'm not opposed to doing something like that. What do others
think? (Right now we just offer a default of 1 for the ADF
lag-length, and leave it up to the user to raise the order, and to
select the option of testing down from some maximum order.)

Dear Allin, today I realized that Gretl are using the Schwert's suggestion to choose
the maximum lag length. Thanks for implementing this!

I would like to suggest to put this information into the Gretl-help. In the fractional
integration test help we have:

"If the optional order argument is not given the order for the test(s) is set automatically
as the lesser of T/2 and T0.6."

So I think it would be nice if we could put something like that into the ADF's help:

"If the lag order argument is not given the order for the test is set automatically as
suggested by Schwert (1989):

Max ADF lag = int(12*($nobs/100)^(1/4))"


Best regards,
Henrique C. de Andrade
Doutorando em Economia Aplicada
Universidade Federal do Rio Grande do Sul
www.ufrgs.br/ppge