[Gretl-users] Features for Gretl 2.0
Hi all!
Reading Allin Cottrells thougtfull paper on extending Gretl and the
importance of the Bundle-Type as Return-Type of functions I noticed how
powerful and similar it is to the Matlab Econometric Toolbox of LeSage
and Grocer, the Econometric Toolbox for Scilab. Since the Syntax for
Function Definition is so similar, it should be easy to translate some
useful and missing functions to Gretl (the code of the Toolboxes is open
and well documented); e.g Grocer entails a function automatic, which
does the general to specific Model Selection of Hendry and Krozlig, only
implemented in PCGets. Well, that is my question for the Developers and
the Gretl-Community: What is useful and/or important, but still missing?
What features should Gretl 2.0 have?
My own wishes and ideas cover broadly two things I miss.
The first, and more important one, concerns what in Gretl is called
system and in Eviews Model. After estimating the system it should be
possible to solve it, make dynamic and stochastic simulations and
forecasts, calculate multipliers, simulate and analyze scenarios etc. I
am not sure, but it seems that one cannot do that in Gretl so far? The
chapter on that in the UserGuide is unwritten. Eviews offers Dialogs for
many options. Ray Fairs Fortran Program Fair-Parke has several more
options (forecast evaluation, optimal control etc), and Fair makes the
source available on his Website (but I cant read Fortran77, anybody out
there?). I think that if Gretl would have such standard and advanced
routines and expose it in the usual user-friendly and nice way, it
would be great (and offer some competitive advantage over EViews).
The second thing I miss is a Database search and query possibility (also
offered by EViews), that is searching and selecting data by some set of
criteria, e.g. Us GDP annual, in current or chained dollar since 1970,
published by BEA. My idea how to implement this is roughly the
following. That criteria arent too many and always roughly the same,
i.e. country, variable, frequency, start-enddate, unit, datasource and
maybe some more. Therefore it would be possible to set up a relational
Database according to a Star Scheme with Dimension Tables for country,
unit etc, so that the complete Information for a Series is given by the
specific combination of entries in the Dimension Tables. Now, for the
Gretl Databases the idx Files already contain most of the Information to
populate the Dimension Tables, so maybe they can be read in in bulk,
adding a few Informations manually. A small Database is SQLite, which
could be easily added since its one sqlite.c file. Then the queries with
sql select combinations could be built and connected to a Gui with e.g.
Comboboxes for each Dimension for selection. Well, does that idea make
sense?
Well, too much for a first statement...but anyway, Gretl is a wonderfull
program and I want to thank Allin for it.