Re: [Gretl-devel] ARMA developments
On Mon, 30 Apr 2018, Sven Schreiber wrote:
Am 30.04.2018 um 01:50 schrieb Allin Cottrell:
> On Sun, 29 Apr 2018, Sven Schreiber wrote:
>
>> Am 29.04.2018 um 21:10 schrieb Allin Cottrell:
>>
>>> up with what I think is now the fastest ARMA around (faster than
>>> x12a/x13a, and much faster than R's arima()).
>>
>> So aim to replace the external X13? Not entirely seriously, but why
>> actually not?
>
> Well, I'd say it's still worth keeping X13 as an optional alternative.
It's
> reasonably fast, and since it uses a quite different algorithm (IGLS) it
> may show up the odd case where our native code misses the MLE.
>
> I used X13 as a baseline in my testing of our new code and it would be a
> pain to have to do that sort of thing externally rather than just via the
> convenient "-x" option to our "arma" command.
Sorry, I wasn't clear, I meant the seasonal adjustment program/algorithm. But
I may well be missing the point, I've never looked at it in detail.
Seasonal adjustment via x13as or tramo/seats has some bells and
whistles that would take time to code ourselves, and besides I think
it's good to offer what's actually used by the statistical agencies.
One could envisage a function package that implemented seasonal
adjustment using our native arma, but I wouldn't rate it a high
priority.
Allin