Re: [Gretl-devel] some 'coint' interface details (and questions)
Am 14.11.18 um 00:12 schrieb Allin Cottrell:
OK, makes good sense. That interpretation of the --seasonals option
for "coint" is now in git. The doc still needs updating; hopefully
that (and revised snapshots) will follow tomorrow.
I have tested the new snapshot and it's in there alright. Impressive how
the new items automatically appears in the local language (though I know
that's because it's referencing existing strings).
There's one remaining glitch AFAICS for the case of no constant
especially. The regular seasonal dummies have a non-zero mean (over
time), so they implicitly introduce a constant (as you know). Thus if
the user chooses no constant, centered seasonal dummies should be used.
It's probably the best way to use centered dummies always.
Plus the --seasonals option isn't honored apparently in the previous ADF
tests when --nc is given, as in this:
<hansl>
open australia
coint 3 IAU IUS --test-down --seasonals --nc # no seas dums in ADF steps
coint 3 IAU IUS --test-down --seasonals # with seas dums
</hansl>
Actually it seems that this is an inherited bug from the ADF routine,
because when I do
<hansl>
open australia
adf 3 IAU --seasonals --nc --verbose
</hansl>
then the seasonals are also ignored. (Of course they must be centered,
as above.)
Maybe this behavior predates the introduction of centered dummies in gretl?
thanks,
sven