Re: [Gretl-users] Mixed Data Sampling Regression
On Tue, 5 Apr 2016, Henrique Andrade wrote:
I think it will be important to have sessions with mixed data. In my
humble
opinion the best design in terms of user experience would be the use of
tabs (like EViews). But I'm almost sure that this is a very difficult
programming task :-(
This would imply such a radical redesign of the internals of gretl that we
might as well start a new project from scratch :)
One thing we could do, however, is setting up a standard for incorporating
a higher-frequency series into a lower-frequency one (the opposite is
quite easy to do). For example, if you have a quarterly dataset, you could
import a monthly series as a "seasonal list" of three variables (one per
subperiod).
For example, suppose you want to pull monthly Industrial Production into a
quarterly dataset. You could have gretl automatically generating 3 new
series "indpro_subp1", "indpro_subp2" and "indpro_subp3",
where
"indpro_subp1" contains January, April, July etc. Not the best, but at
least
* no information is lost
* operations like lagging, computing moving averages etc are trickier but
possible by operating on the list instead of the single series
* converting the dataset to higher frequency becomes easy
The only problem would arise with periodicities that are not strict
multiple of one another; right now, the only example that springs to my
mind is weekly data in monthly/quarterly datasets. Any other?
-------------------------------------------------------
Riccardo (Jack) Lucchetti
Dipartimento di Scienze Economiche e Sociali (DiSES)
Università Politecnica delle Marche
(formerly known as Università di Ancona)
r.lucchetti(a)univpm.it
http://www2.econ.univpm.it/servizi/hpp/lucchetti
-------------------------------------------------------