[Gretl-users] Re: advice on function package: to share or not to share
On Wed, 3 Feb 2021, Artur Bala wrote:
Hi all,
Sometimes ago I developed a script, and eventually a function package,
that replicates a built-in Stata command for outliers-corrected
estimation (rreg) (basically iterative WLS).
Well, digging further in the literature (and forums) I discover that the
method behind this Stata command, though widely used, becomes
ineffective under certain assumptions (eg. clusters of outliers) and new
more effective estimation procedures are currently available in the
literature (and also implemented in some user-provided Stata commands).
Sure I don’t regret the hours spent on “my” function but I wonder if
it’s still worthy to share it with the community since I know that it
does not fit to all situations.
For what it's worth, my advice is "go ahead with what you have". Even if
the method you implemented is not the fanciest, some people still may find
it useful, if only for replicating old results.
Of course, it would be nice to
(a) state the limitations of the method you coded clearly in the
documentation
(b) implement the more recent alternatives, possibly as part of the same
function package.
-------------------------------------------------------
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
-------------------------------------------------------